Tellers

43-3071.00
Median wage $43,030/yr329,480 employed (US)Rank #17 of 923 scored · top 2% by substitution

Receive and pay out money. Keep records of money and negotiable instruments involved in a financial institution's various transactions.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution69
Exposure70
Augmentation63

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

28 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

57%

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%70

panel mean rating 3.8/5 → substitution pressure 70/100

Technical feasibility todayw 20%70

panel mean rating 3.8/5 → substitution pressure 70/100

Cost vs. human wagew 15%78

panel mean rating 4.1/5 → substitution pressure 78/100

Adoption barriersw 20%inverted — strong barriers lower the score57

panel mean rating 2.7/5 (barrier strength) → substitution pressure 57/100

Sector adoption velocityw 10%70

panel mean rating 3.8/5 → substitution pressure 70/100

Task breakdown (28 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.

Quote unit exchange rates, following daily international rate sheets or computer displays.

97

CI 95–100 · exposure 100 · augmentation 63 · importance 3.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Banking and fintech sectors have rapidly adopted automated exchange rate quoting across digital channels; self-service platforms, mobile apps, and online banking systems now provide instant rate quotes without human intervention.
Sector adoption velocityclaude-sonnet-55/5Banking and financial services have rapidly and broadly adopted digital rate displays and self-service kiosks, largely displacing manual rate quoting.
Augmentation potentialclaude-haiku-4-5-202510014/5AI assistants can augment tellers by providing real-time rate lookups, flagging anomalies, or suggesting competitive options, allowing human tellers to focus on customer relationship and complex transaction aspects while staying informed.
Augmentation potentialclaude-sonnet-53/5Where tellers still exist, computer displays already assist them by providing instant rate lookups, though this is a narrow slice of their overall duties.
Task automatabilityclaude-haiku-4-5-202510015/5Quoting exchange rates is straightforward data retrieval and transmission from published sources or system displays; current AI systems can reliably access, verify, and deliver accurate rates with minimal human time investment, easily meeting the 50% time-saving threshold.
Task automatabilityclaude-sonnet-55/5Looking up and quoting a rate from a rate sheet or computer display is a trivial data-retrieval task that automated systems (kiosks, apps, APIs) already perform instantly and accurately.'
Adoption barriersclaude-haiku-4-5-202510012/5While regulatory frameworks govern currency exchange compliance, the task of quoting published rates has few hard barriers to automation; most organizations simply retain human tellers for customer service and other tasks rather than legal prohibition.
Adoption barriersclaude-sonnet-51/5There is no licensing or legal requirement that a human teller personally quote exchange rates; it's pure information display with no liability concerns.
Cost vs. human wageclaude-haiku-4-5-202510015/5The AI cost of retrieving and transmitting an exchange rate quote (API call, minimal inference) is orders of magnitude lower than the loaded wage of a human teller performing the same function.
Cost vs. human wageclaude-sonnet-55/5Displaying a rate from a data feed costs fractions of a cent via automated systems versus paying a human teller's wage for the same lookup.
Technical feasibility todayclaude-haiku-4-5-202510015/5Exchange rate APIs and automated quotation systems are mature, deployed products in banking and financial services; many institutions already automate rate delivery to customers through ATMs, web, and mobile interfaces.
Technical feasibility todayclaude-sonnet-55/5Currency exchange rate quoting is already fully deployed via ATMs, banking apps, and online converters used at scale in production.

Compute financial fees, interest, and service charges.

95

CI 95–95 · exposure 100 · augmentation 75 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Banking and financial services are among the earliest and deepest adopters of automation; fee and interest calculation has been computerized for 40+ years and is now standard across all digital banking platforms.
Sector adoption velocityclaude-sonnet-55/5Retail banking has already deeply and rapidly automated this specific sub-task; virtually no institution manually computes these figures today.
Augmentation potentialclaude-haiku-4-5-202510014/5While tellers may still review or apply calculated fees in context, AI computation directly supports and augments their workflow by instantly providing accurate fee figures, reducing manual arithmetic and allowing focus on customer service and exception-handling.
Augmentation potentialclaude-sonnet-54/5Where tellers still verify or explain calculated charges to customers, automated computation tools significantly speed their workflow and reduce errors.
Task automatabilityclaude-haiku-4-5-202510015/5Computing financial fees, interest, and service charges involves straightforward mathematical calculations based on predefined rules and formulas. Current AI systems (including spreadsheet automation and banking software) can execute these calculations end-to-end with 100% accuracy and near-zero time cost, far exceeding the 50% time-saving threshold.
Task automatabilityclaude-sonnet-55/5Computing fees, interest, and service charges is a deterministic, rules-based calculation already fully handled by core banking software and calculators, requiring no human judgment.
Adoption barriersclaude-haiku-4-5-202510012/5While banking is regulated, the calculation itself is not restricted to licensed humans—any system can compute fees and interest as long as the underlying rules comply with regulation. Oversight is mainly rule-validation and audit, not a human sign-off requirement on each calculation.
Adoption barriersclaude-sonnet-52/5No licensing requirement blocks automated calculation itself, though bank policies may require teller verification or dual sign-off on certain overrides, creating minor friction.
Cost vs. human wageclaude-haiku-4-5-202510015/5AI computation cost for this task is negligible (fractions of a cent per calculation), while human teller labor costs $15–25+ per hour loaded; the cost ratio is easily orders of magnitude in AI's favor.
Cost vs. human wageclaude-sonnet-55/5Automated calculation via software costs fractions of a cent per transaction versus paying a human teller's loaded wage to compute the same figures manually.
Technical feasibility todayclaude-haiku-4-5-202510015/5Banking and fintech products have reliably automated fee and interest calculation for decades; it is core functionality in every major banking platform, loan servicing system, and accounting software deployed at scale in production across millions of accounts.
Technical feasibility todayclaude-sonnet-55/5Banking systems have automated fee/interest computation in production for decades; this is standard functionality in every deployed teller and core banking platform.

Sort and file deposit slips and checks.

94

CI 92–95 · exposure 100 · augmentation 38 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Banking and financial services are digitization leaders with early, deep adoption of automation infrastructure. Check processing automation has been standard practice at major institutions for 15+ years.
Sector adoption velocityclaude-sonnet-54/5Banking is a highly digitized industry with widespread deployment of automated deposit and check-processing technology, though small branches may retain manual steps.
Augmentation potentialclaude-haiku-4-5-202510012/5This task offers minimal augmentation opportunity since the sorted/filed output is already a finished product; AI assists mainly by removing the need for human involvement rather than enhancing human capability in a hybrid workflow.
Augmentation potentialclaude-sonnet-53/5Where manual sorting still occurs, scanning and workflow tools assist tellers by reducing physical handling and error, though the task is largely already automated rather than merely augmented.
Task automatabilityclaude-haiku-4-5-202510015/5Sorting and filing deposit slips and checks is a highly repetitive, rule-based task involving image capture (check scanning), optical character recognition (OCR), and structured data routing. Current banking systems already automate this end-to-end with >50% time savings through check image processing and automated filing workflows.
Task automatabilityclaude-sonnet-55/5Sorting and filing deposit slips and checks is a routine, structured, rules-based clerical task easily handled by document imaging, OCR, and automated check-processing systems already in wide use.
Adoption barriersclaude-haiku-4-5-202510012/5Minimal regulatory barriers exist for automating back-office sorting and filing; banks already do this. Some organizational friction may remain around legacy systems or transition costs, but no licensing or human-sign-off requirement applies to the sorting task itself.
Adoption barriersclaude-sonnet-52/5No licensing requirement attaches to sorting/filing; some internal audit and reconciliation controls exist but do not require a human to physically perform this specific step.
Cost vs. human wageclaude-haiku-4-5-202510015/5Automated check processing infrastructure (scanners, OCR, routing) has already achieved economies of scale in the banking industry; per-item costs are orders of magnitude lower than manual sorting and filing labor.
Cost vs. human wageclaude-sonnet-55/5Automated check-scanning and filing systems process thousands of items per hour at a fraction of the cost of manual teller sorting labor.
Technical feasibility todayclaude-haiku-4-5-202510015/5Production-grade check scanning and automated document routing systems are mature and deployed at scale in banking institutions. Systems like Federal Reserve-compliant check imaging and bank-integrated document management platforms perform this reliably daily.
Technical feasibility todayclaude-sonnet-55/5Bank check imaging, Remote Deposit Capture, and automated sorting/reconciliation systems have been standard production technology in banking for over a decade.

Enter customers' transactions into computers to record transactions and issue computer-generated receipts.

89

CI 84–95 · exposure 92 · augmentation 50 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5The banking and financial services sector has already heavily adopted automation for this task (ATMs since the 1970s, online banking, mobile payment apps); teller transaction processing is one of the longest-standing examples of successful AI/automation displacement in production.
Sector adoption velocityclaude-sonnet-54/5Banking is a heavily digitized sector with fast, deep adoption of self-service and automated transaction channels, already displacing substantial teller transaction volume industry-wide.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist tellers by auto-populating fields, suggesting transaction types, and flagging errors in real-time, improving speed and accuracy. However, for routine transactions, the teller's role is increasingly peripheral rather than enhanced, since the system itself handles the core entry and receipt generation.
Augmentation potentialclaude-sonnet-53/5Where human tellers remain, computer systems already streamline transaction entry and receipt generation, though this is more full substitution than augmentation for the core data-entry action itself.
Task automatabilityclaude-haiku-4-5-202510015/5This task is purely data entry and receipt generation—fully automatable end-to-end via ATMs, online banking, and point-of-sale systems that already achieve far >50% time savings by eliminating manual keystroke and paper receipt steps. Modern banking systems handle transaction recording and receipt issuance automatically.
Task automatabilityclaude-sonnet-54/5Data entry of standardized transactions into a computer system is highly structured and repetitive, well within scope of automation via ATMs, ITMs, and online banking interfaces that already replace this exact function for most transaction types.
Adoption barriersclaude-haiku-4-5-202510012/5While banks face some customer preference for human interaction and must maintain compliance/audit trails, there are no licensing requirements mandating a human teller perform transaction recording. Regulatory frameworks support automated processing, and organizations face little legal or liability friction in automating this narrow task.
Adoption barriersclaude-sonnet-52/5No licensing requirement exists for basic transaction entry, though some customers still prefer human tellers and banks maintain branch staffing for complex or high-value transactions and regulatory recordkeeping oversight.
Cost vs. human wageclaude-haiku-4-5-202510015/5The marginal cost of processing one additional digital transaction is near-zero once systems are in place, whereas a human teller's loaded wage is substantial; AI-driven transaction processing is orders of magnitude cheaper per transaction.
Cost vs. human wageclaude-sonnet-55/5Automated transaction processing via ATM/digital banking infrastructure costs a small fraction per transaction compared to a loaded teller wage for the same interaction.
Technical feasibility todayclaude-haiku-4-5-202510015/5Deployed systems (ATMs, mobile banking, online portals, and teller-assist software) reliably perform transaction recording and receipt generation in production across the financial sector at massive scale with negligible error rates on routine transactions.
Technical feasibility todayclaude-sonnet-55/5ATMs, mobile banking apps, and interactive teller machines already perform this task reliably at massive scale in production across virtually all major banks.

Count currency, coins, and checks received, by hand or using currency-counting machine, to prepare them for deposit or shipment to branch banks or the Federal Reserve Bank.

87

CI 80–95 · exposure 92 · augmentation 75 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Banking and financial services are among the most digitized sectors with early and sustained adoption of automation. Currency-counting machines are standard infrastructure in branches nationwide, representing deep, mature adoption.
Sector adoption velocityclaude-sonnet-54/5Banking is a highly digitized sector and currency-counting machines are already standard equipment in most branches and cash centers.
Augmentation potentialclaude-haiku-4-5-202510014/5Automated counting machines augment teller productivity significantly by offloading the mechanical counting task, allowing tellers to focus on customer service and verification, while the human remains responsible for deposit accuracy oversight.
Augmentation potentialclaude-sonnet-54/5Counting machines substantially speed up and improve accuracy of the teller's counting work while the teller still verifies totals and handles exceptions.
Task automatabilityclaude-haiku-4-5-202510015/5Currency and coin counting is highly structured, repetitive work with clear numeric outputs. Modern currency-counting machines and AI-enabled systems can process and verify deposits end-to-end with >50% time savings and equal or better accuracy than manual hand counting.
Task automatabilityclaude-sonnet-54/5Currency-counting machines and cash recyclers already automate the counting and sorting portion end-to-end with high speed and accuracy, though physical handling and exception resolution still require a person.
Adoption barriersclaude-haiku-4-5-202510012/5While banks have strong internal controls and audit requirements, these apply to the process outcome (accuracy verification) rather than prohibiting automation. However, regulatory oversight and error-cost sensitivity for financial institutions create modest friction in deployment.
Adoption barriersclaude-sonnet-52/5No licensing requirement blocks machine counting, but banks maintain human oversight for reconciliation, fraud detection, and dual-control policies around cash handling.
Cost vs. human wageclaude-haiku-4-5-202510015/5Automated currency-counting machines have very low per-transaction operational costs compared to the fully loaded wage of a teller performing repetitive counting work, typically amortizing over thousands of transactions daily.
Cost vs. human wageclaude-sonnet-54/5Machine counting is far faster and cheaper per transaction than manual counting once the capital equipment is in place, though upfront hardware cost and maintenance keep it from a full order-of-magnitude advantage everywhere.
Technical feasibility todayclaude-haiku-4-5-202510015/5Currency-counting machines are mature, widely deployed products in banking that reliably perform this task at scale. Many banks have already automated much of this work with commercially available hardware and software solutions.
Technical feasibility todayclaude-sonnet-55/5Currency-counting and coin-sorting machines are mature, widely deployed products used in banks and cash-handling operations at scale today.

Receive checks and cash for deposit, verify amounts, and check accuracy of deposit slips.

87

CI 79–95 · exposure 92 · augmentation 50 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Banking and financial services are highly digitized and early adopters of AI; mobile check deposit and back-end automation are standard across major institutions, though some branches still retain manual teller roles for customer service and exceptions.
Sector adoption velocityclaude-sonnet-55/5Retail banking has aggressively adopted ATMs, mobile deposit, and self-service kiosks for decades, representing deep, fast, mature adoption of this specific task.
Augmentation potentialclaude-haiku-4-5-202510013/5AI assists tellers by pre-scanning checks, suggesting amounts, and flagging discrepancies, reducing manual verification time and errors without full replacement—useful on the routine parts of the task but tellers retain final sign-off responsibility.
Augmentation potentialclaude-sonnet-53/5For transactions still handled by humans, AI-assisted counting machines and verification software speed up the process, though the task is largely replaced rather than augmented.
Task automatabilityclaude-haiku-4-5-202510014/5Bank tellers' deposit verification can be largely automated by image recognition, OCR, and automated cash counting. Mobile check deposit apps already perform check scanning and amount verification; remaining manual steps (cash counting, slip verification) are simple enough for current systems to handle with >50% time savings at equal quality.
Task automatabilityclaude-sonnet-55/5Deposit verification and amount checking is already fully automated via ATMs, ITMs, and mobile deposit capture with OCR/check-reading AI, meeting the 50% time-saving bar easily since it removes the human step entirely for most transactions.
Adoption barriersclaude-haiku-4-5-202510013/5Regulatory oversight (AML, Know Your Customer, Dodd-Frank) requires human verification or sign-off on some deposits; customer service expectations and occasional fraud exceptions also create friction. However, the core deposit-slip accuracy check is unencumbered.
Adoption barriersclaude-sonnet-52/5No licensing requirement for this specific task; some friction exists from customers preferring in-person service and banks needing fraud/error controls, but no legal requirement for human handling.
Cost vs. human wageclaude-haiku-4-5-202510015/5Automated check imaging, OCR, and cash-counting systems cost pennies per transaction, far cheaper than a human teller's loaded wage (~$32k/year) divided across hundreds of daily deposits.
Cost vs. human wageclaude-sonnet-55/5Automated deposit processing via machines/apps costs a small fraction of a teller's loaded wage per transaction once infrastructure is in place.
Technical feasibility todayclaude-haiku-4-5-202510015/5Deployed products perform check deposit verification and amount validation reliably in production today—mobile check deposit is ubiquitous across banks, and automated teller cash handling systems are mature and widely deployed in banking infrastructure.
Technical feasibility todayclaude-sonnet-55/5Bank ATMs, remote deposit capture apps, and image-based check recognition are mature, widely deployed production systems used by virtually all major banks today.

Identify transaction mistakes when debits and credits do not balance.

87

CI 79–95 · exposure 87 · augmentation 75 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Banks are early and aggressive adopters of fintech automation. Reconciliation and transaction-error detection are core backend processes actively automated across major financial institutions.
Sector adoption velocityclaude-sonnet-54/5Financial services is a fast-adopting sector for automation of back-office and transaction-processing functions, with reconciliation automation widely deployed.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can surface flagged discrepancies and provide automated highlighting of mismatched entries, significantly reducing human review time and error detection effort even when humans remain responsible for final sign-off.
Augmentation potentialclaude-sonnet-54/5AI-driven flagging tools substantially speed up a teller's or back-office worker's ability to spot and address discrepancies, letting humans focus on judgment-heavy resolution.
Task automatabilityclaude-haiku-4-5-202510015/5Reconciling debits and credits is a deterministic arithmetic task with clear success criteria. Current AI systems can reliably identify imbalances by comparing numerical records end-to-end, easily exceeding the 50% time-saving threshold.
Task automatabilityclaude-sonnet-54/5Reconciliation and balance-checking is a rules-based numerical task well-suited to software; automated reconciliation systems can flag mismatches with minimal human input, saving significant time. Full end-to-end automation exists but human tellers may still need to resolve edge cases.
Adoption barriersclaude-haiku-4-5-202510012/5While some banks may require human sign-off for final reconciliation due to risk management preferences, the task itself has no legal licensing requirement and automation is widely permitted and deployed.
Adoption barriersclaude-sonnet-52/5Banking has regulatory oversight and audit requirements, but the specific act of flagging an imbalance is not legally reserved to a licensed human; some institutional friction exists around sign-off on corrections.
Cost vs. human wageclaude-haiku-4-5-202510015/5Inference costs for arithmetic validation are trivial (fractions of a cent per transaction), while teller time costs $15–25/hour loaded; the cost ratio is orders of magnitude in AI's favor.
Cost vs. human wageclaude-sonnet-55/5Automated ledger-balancing and anomaly-detection software runs at near-zero marginal cost per transaction compared to a teller's wage for manual reconciliation.
Technical feasibility todayclaude-haiku-4-5-202510015/5Bank reconciliation and transaction validation systems are deployed at scale across the financial industry. Automated reconciliation software and AI-driven anomaly detection reliably perform this task in production today.
Technical feasibility todayclaude-sonnet-54/5Banking core systems and reconciliation software already flag debit/credit imbalances automatically in production at scale, though some discrepancies still require human review for root-cause investigation.

Receive mortgage, loan, or public utility bill payments, verifying payment dates and amounts due.

85

CI 84–86 · exposure 84 · augmentation 63 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Banking and fintech sectors have been early, aggressive adopters of automated payment processing; mobile banking, ACH automation, and online bill pay have displaced manual teller-based payment receipt at scale for over a decade.
Sector adoption velocityclaude-sonnet-54/5Retail banking has aggressively shifted routine transactions to online/mobile/ATM channels for over a decade, representing fast, deep adoption of automation for this specific task type.
Augmentation potentialclaude-haiku-4-5-202510014/5AI tools like payment dashboards, automated exception flagging, and OCR-assisted verification significantly augment human tellers when they do handle complex or flagged cases, improving their accuracy and speed on the residual exceptions.
Augmentation potentialclaude-sonnet-53/5For transactions still handled in person, AI-backed systems help tellers verify amounts and flag discrepancies faster, though the task itself is largely already automated outside human workflow.
Task automatabilityclaude-haiku-4-5-202510014/5Payment receipt, verification of dates, and amount validation are highly structured, rule-based tasks that AI systems can automate nearly end-to-end. OCR and regex-based matching on bills, combined with database lookup, easily achieves 50%+ time savings; the main residual is exception handling and customer clarification, which is a small fraction of routine transactions.
Task automatabilityclaude-sonnet-54/5Verifying payment dates/amounts and processing routine bill or loan payments is a structured, rules-based data task that automated systems (online banking, kiosks, payment processors) already handle at scale with minimal human involvement.
Adoption barriersclaude-haiku-4-5-202510012/5While financial services face compliance oversight (BSA, AML regulations), payment receipt and verification itself does not require a licensed human signatory; most barriers are organizational (customer preference for human contact, legacy system integration) rather than legal prohibition. Regulatory requirements apply to the process but not to who performs verification.
Adoption barriersclaude-sonnet-52/5Some regulatory requirements around fraud verification and dispute handling exist, but no licensing requirement mandates a human teller for routine payment receipt, and many payments already bypass tellers entirely.
Cost vs. human wageclaude-haiku-4-5-202510015/5AI-driven payment processing (machine learning-based OCR, rule engines, and automated reconciliation) costs a fraction of a cent per transaction, while a human teller's loaded wage is $25–35 per hour; the cost ratio is orders of magnitude in AI's favor.
Cost vs. human wageclaude-sonnet-55/5Automated payment processing costs fractions of a cent per transaction versus a loaded teller wage, an order-of-magnitude or greater cost advantage.
Technical feasibility todayclaude-haiku-4-5-202510015/5Payment processing and bill verification are mature, deployed capabilities in banking and fintech systems today. Automated payment platforms, mobile banking apps, and back-office OCR already reliably perform this task at scale in production across thousands of institutions.
Technical feasibility todayclaude-sonnet-55/5Automated bill pay, ATMs, online banking portals, and payment processing systems reliably perform this exact function in production for millions of transactions daily.

Balance currency, coin, and checks in cash drawers at ends of shifts and calculate daily transactions, using computers, calculators, or adding machines.

84

CI 72–95 · exposure 87 · augmentation 75 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Banks and financial institutions have been automating cash reconciliation and transaction calculation for decades; modern teller workstations are already heavily integrated with automated systems that perform these calculations, with widespread production deployment across the sector.
Sector adoption velocityclaude-sonnet-53/5Banking has moderate-to-strong digitization, with many branches adopting cash recycler/ATM-integrated systems, but full automation varies widely by institution size and legacy infrastructure.
Augmentation potentialclaude-haiku-4-5-202510014/5AI systems augment tellers by automating the calculation and initial reconciliation steps, allowing humans to focus on discrepancy investigation and exception handling; this significantly raises productivity while the teller retains oversight and final verification responsibility.
Augmentation potentialclaude-sonnet-54/5Calculators, computers, and automated counting machines already substantially speed up and reduce errors in this balancing task while tellers remain responsible for final verification.
Task automatabilityclaude-haiku-4-5-202510015/5Balancing cash drawers and calculating daily transactions are straightforward arithmetic tasks with clear inputs (currency counts, checks, transaction records) and outputs (balanced totals). Modern accounting software and AI systems can reliably perform these calculations end-to-end with 100% accuracy and significant time savings compared to manual entry.
Task automatabilityclaude-sonnet-54/5Cash drawer reconciliation is a structured numerical task already handled by cash-counting machines and POS/teller systems that auto-tally transactions, though physical cash counting still requires some human or hardware handling.4
Adoption barriersclaude-haiku-4-5-202510012/5While banks often require human sign-off on final cash counts for audit and accountability purposes, the actual computational work of balancing and reconciling transactions faces minimal regulatory or legal barriers to automation; oversight is mainly organizational practice rather than legal mandate.
Adoption barriersclaude-sonnet-52/5Some internal audit and dual-control policies exist for cash handling, but no licensing requirement mandates a human perform this specific reconciliation task.
Cost vs. human wageclaude-haiku-4-5-202510015/5The per-transaction cost of automated calculation and reconciliation is orders of magnitude cheaper than paying a teller's loaded wage ($15–20/hour) to manually count and verify cash, especially when amortized across thousands of transactions.
Cost vs. human wageclaude-sonnet-54/5Automated cash counters and reconciliation software are relatively inexpensive to operate compared to the labor time spent manually balancing drawers, though hardware capital costs exist.
Technical feasibility todayclaude-haiku-4-5-202510015/5Deployed banking and accounting systems (including Excel-based tools, QuickBooks, and bank-integrated software) already perform transaction reconciliation and cash drawer balancing routinely in production at scale across financial institutions.
Technical feasibility todayclaude-sonnet-54/5Banks widely deploy teller cash recyclers and automated balancing software that reconcile drawers with minimal manual calculation, though full end-to-end automation without any human check is less common in smaller branches.

Arrange monies received in cash boxes and coin dispensers according to denomination.

81

CI 70–92 · exposure 87 · augmentation 38 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Major banks and high-volume retail operations have already widely deployed automated coin sorters and cash-handling systems; adoption is mainstream in digitized financial environments.
Sector adoption velocityclaude-sonnet-54/5Banking has adopted cash recycling and coin-counting automation extensively for decades, though small branches and small retailers still do this manually.
Augmentation potentialclaude-haiku-4-5-202510012/5Once a teller delegates money arrangement to a machine, there is limited scope for AI to augment their participation in the task itself; the task is primarily mechanical counting and organizing rather than judgment-based.
Augmentation potentialclaude-sonnet-53/5Where manual sorting persists, coin counting machines and cash drawers with denominated slots assist tellers but don't transform judgment-based work.
Task automatabilityclaude-haiku-4-5-202510014/5A physical robotic or coin-sorting system can fully automate arranging money by denomination end-to-end. While current AI doesn't inherently solve the mechanical problem, integrated coin-sorting and cash-handling machines already exist and deliver well over 50% time savings versus manual counting and arranging.
Task automatabilityclaude-sonnet-55/5This is a mechanical sorting/organizing task already fully automated by cash recyclers and coin sorting machines widely deployed in banks and retail.
Adoption barriersclaude-haiku-4-5-202510013/5While there are no strict licensing barriers, deployment requires significant capital investment, physical space, and integration with existing cash-handling procedures, creating organizational friction that slows but does not prevent adoption.
Adoption barriersclaude-sonnet-52/5No licensing requirement for cash sorting itself, though some internal audit/reconciliation controls create minor procedural friction.
Cost vs. human wageclaude-haiku-4-5-202510014/5Modern automated coin and cash dispensers cost several thousand dollars upfront but process hundreds of transactions per day with minimal labor, making per-task cost dramatically lower than the loaded wage of a teller performing this manually.
Cost vs. human wageclaude-sonnet-55/5Automated cash-handling machines process and sort currency far faster and cheaper per transaction than manual teller sorting once amortized.
Technical feasibility todayclaude-haiku-4-5-202510014/5Mature coin sorters and cash-handling machines are deployed in banks and retail environments today, reliably sorting and arranging currency by denomination at scale with low error rates.
Technical feasibility todayclaude-sonnet-55/5Cash recycling machines, coin counters, and automated teller cash dispensers reliably perform this exact function in production at scale across banks and retailers today.

Process transactions, such as term deposits, retirement savings plan contributions, automated teller transactions, night deposits, and mail deposits.

80

CI 74–86 · exposure 80 · augmentation 50 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Banking and financial services have among the fastest adoption of automation; transaction processing automation is already widespread (ATMs, online banking, mobile deposits). Teller roles are actively shrinking as transactions shift to digital channels.
Sector adoption velocityclaude-sonnet-54/5Retail banking has aggressively adopted automation (ATMs, mobile deposit, online banking) for decades, representing one of the more digitized and automation-forward sectors of consumer finance.
Augmentation potentialclaude-haiku-4-5-202510013/5AI assists human tellers by auto-categorizing transactions, flagging anomalies, and pre-filling forms, but the core task itself is increasingly automated rather than augmented; human tellers now focus on customer service and complex issues rather than transaction processing per se.
Augmentation potentialclaude-sonnet-53/5For the residual human-handled transactions, AI-assisted systems (fraud flags, data entry validation, customer verification tools) can speed up teller workflows, though the core transaction logic is already automated rather than augmented.
Task automatabilityclaude-haiku-4-5-202510014/5Most transaction types (deposits, withdrawals, transfers) can be routed and processed with minimal human intervention via existing banking automation systems. However, edge cases (fraud checks, customer disputes, complex account reconciliations) still require human judgment, preventing a full 5.
Task automatabilityclaude-sonnet-54/5Transaction processing is highly structured and rule-based, and most of it (ATM, night deposit, mail deposit, standard account transactions) is already handled by automated banking systems with minimal human involvement in the actual processing step.
Adoption barriersclaude-haiku-4-5-202510012/5Regulatory oversight applies to transaction accuracy and compliance (AML/KYC), but these are addressed through system controls rather than requiring human sign-off. Customer service and deposit verification can create friction, but many transactions are already fully automated with minimal barriers.
Adoption barriersclaude-sonnet-53/5Banking transactions are regulated and require audit trails, KYC/AML compliance, and error accountability, but the processing itself is not legally required to be performed by a human, only overseen by licensed institutions.
Cost vs. human wageclaude-haiku-4-5-202510015/5Automated transaction processing is orders of magnitude cheaper than human tellers when accounting for infrastructure spread across thousands of transactions; a single teller costs $30k–$40k annually, while AI transaction routing costs pennies per transaction.
Cost vs. human wageclaude-sonnet-55/5Automated transaction processing systems cost fractions of a cent per transaction versus a loaded teller wage, making software-based processing dramatically cheaper at scale.
Technical feasibility todayclaude-haiku-4-5-202510015/5Banks have deployed automated transaction processing at scale for decades; ATMs, night deposit systems, and mail deposit scanning all perform these functions reliably in production today. Modern APIs and RPA further automate routing and record-keeping.
Technical feasibility todayclaude-sonnet-54/5Banks already run these transaction types through automated cores, ATMs, and back-office processing systems at scale in production; teller involvement is increasingly limited to exception handling and customer-facing verification.

Perform clerical tasks, such as typing, filing, and microfilm photography.

80

CI 67–92 · exposure 78 · augmentation 63 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Financial services and banking sectors have rapidly deployed RPA and document automation systems for clerical work; this is among the earliest and deepest automation patterns in the industry, with widespread production use.
Sector adoption velocityclaude-sonnet-53/5Banking has moderate-to-fast digitization of back-office clerical work, but teller roles specifically remain a mix of legacy and modern systems, so this narrow task shows middling adoption.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist humans with these tasks by suggesting filing classifications or accelerating data entry, but since the tasks are routine and highly automatable, augmentation is less salient than replacement.
Augmentation potentialclaude-sonnet-54/5AI-powered document management, transcription, and filing tools meaningfully boost clerical productivity for tellers performing these support tasks, even where full automation is incomplete.
Task automatabilityclaude-haiku-4-5-202510015/5Typing, filing, and microfilm photography are well-defined clerical tasks that modern AI and RPA systems can perform end-to-end with significant time savings. Document processing, data entry automation, and image digitization are mature capabilities that easily exceed the 50% time-saving threshold.
Task automatabilityclaude-sonnet-54/5Typing and filing are largely digital document handling and organization tasks that current AI/software can perform with substantial time savings; microfilm photography is a legacy physical process that resists automation but is a minor component.
Adoption barriersclaude-haiku-4-5-202510012/5Minimal legal barriers exist for automating clerical tasks like typing and filing. Some organizational friction may arise from legacy systems or human-preference policies, but nothing prevents substitution of these administrative functions.
Adoption barriersclaude-sonnet-51/5Clerical tasks like typing and filing carry no licensing or liability requirements and are routinely automated in banking operations without regulatory obstruction.
Cost vs. human wageclaude-haiku-4-5-202510015/5RPA and AI-driven document processing are orders of magnitude cheaper than human clerical labor when amortized across volume. Inference, integration, and oversight costs are minimal compared to loaded wages for tellers performing these tasks.
Cost vs. human wageclaude-sonnet-54/5Automated typing and digital filing systems are far cheaper per unit of output than manual clerical labor once implemented, though legacy microfilm equipment costs somewhat offset savings.
Technical feasibility todayclaude-haiku-4-5-202510014/5Deployed products (RPA platforms, OCR systems, document management software) reliably handle typing, filing, and microfilm digitization in production environments. While occasional edge cases exist, these tasks are routinely automated in banking and financial institutions at scale.
Technical feasibility todayclaude-sonnet-53/5Document management systems, OCR, and AI-assisted filing/typing tools are deployed in banking back-offices, but microfilm-specific processes remain manual and niche, limiting overall reliability across the full task.

Receive and count daily inventories of cash, drafts, and travelers' checks.

74

CI 62–86 · exposure 80 · augmentation 75 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Banks and financial institutions have deployed automated currency-counting and check-scanning systems at scale for decades; adoption is already deep and mature across the sector, reflecting high digitization and capital investment in automation.
Sector adoption velocityclaude-sonnet-53/5Many mid-size and large banks have adopted cash recyclers and automated counting, but smaller branches and credit unions still rely heavily on manual counting, giving mixed penetration.
Augmentation potentialclaude-haiku-4-5-202510014/5AI-powered scanning and counting systems actively assist tellers by automating the tedious manual counting component, displaying results in real-time, and flagging discrepancies; this augmentation significantly increases teller speed and accuracy while they oversee and reconcile the process.
Augmentation potentialclaude-sonnet-54/5Even where humans still count, machine counters and reconciliation software substantially speed up and reduce errors in the verification process, keeping the teller in a supervisory role.
Task automatabilityclaude-haiku-4-5-202510014/5Cash counting and inventory of standardized items like drafts and travelers' checks can be largely automated with high-speed counting machines and optical scanning systems; the core workflow—receiving, sorting, and tallying—is highly repetitive and rule-based, meeting significant time-saving and equal-quality thresholds, though final reconciliation may require human oversight.
Task automatabilityclaude-sonnet-54/5Cash counting and reconciliation is highly structured and already handled by cash-recycling machines and automated teller safes/ATMs that count, validate, and log currency with minimal human input.ureaucratic paperwork. Full end-to-end automation exists for most of this task, though drafts/travelers checks add slight manual verification.
Adoption barriersclaude-haiku-4-5-202510014/5Banking regulations and reconciliation requirements impose mandatory audit trails and human sign-off on cash counts; compliance, liability, and fiduciary responsibility create strong organizational and legal friction against full unsupervised automation.
Adoption barriersclaude-sonnet-52/5Some internal audit/control policies require dual verification or human sign-off for cash discrepancies, but no licensing law mandates a human count cash himself.
Cost vs. human wageclaude-haiku-4-5-202510015/5Automated counting systems and optical scanners cost a fixed amount per deployment and scale across hundreds of transactions daily, whereas a human teller performs this task once per shift; the per-transaction cost of automation is orders of magnitude lower.
Cost vs. human wageclaude-sonnet-54/5Once installed, cash-counting machines process large volumes far faster and cheaper per transaction than manual counting, though upfront hardware costs are non-trivial compared to pure software AI.
Technical feasibility todayclaude-haiku-4-5-202510015/5Deployed currency-counting machines and check-scanning systems are mature, production-grade technologies widely used in banking and retail; these systems reliably handle high volumes with documented error rates acceptable to financial institutions.
Technical feasibility todayclaude-sonnet-54/5Bank-grade cash counters, smart safes, and ATM/TCR (teller cash recycler) systems are widely deployed in production at scale in retail banking, reliably counting and reconciling cash drawers daily.

Compose, type, and mail customer statements and other correspondence related to issues such as discrepancies and outstanding unpaid items.

73

CI 67–79 · exposure 70 · augmentation 63 · importance 3.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Banking and financial services are digitally mature sectors with strong incentives to automate high-volume, repetitive customer communication; many banks already use automated statement and correspondence systems.
Sector adoption velocityclaude-sonnet-53/5Banking has moderate AI adoption for back-office document automation, though tellers themselves are in a slower-adopting, branch-based segment of the industry.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by auto-drafting letters and flagging discrepancies, allowing a teller to review and personalize rather than compose from scratch, raising throughput for routine cases.
Augmentation potentialclaude-sonnet-54/5AI drafting tools can significantly speed up composing routine correspondence, letting tellers review and personalize rather than write from scratch.
Task automatabilityclaude-haiku-4-5-202510014/5Most of this task can be automated end-to-end: template-based statement generation, identification of discrepancies from transaction logs, automated letter composition, and mail merging are all standard capabilities. A human may still need to review complex cases, but routine high-volume correspondence easily meets the ≥50% time-saving threshold.
Task automatabilityclaude-sonnet-54/5Drafting standardized correspondence about discrepancies and unpaid items is largely templated text generation that LLMs handle well, though identifying the specific account issue requires system data integration.dd
Adoption barriersclaude-haiku-4-5-202510012/5Banks can legally automate routine correspondence and statements without human sign-off; there are no licensing or hard regulatory barriers to substitution. Some internal oversight policies may exist, but they are not hard legal barriers.
Adoption barriersclaude-sonnet-52/5Some compliance review may be needed for financial correspondence, but this is largely administrative rather than requiring licensed sign-off, so barriers are modest.
Cost vs. human wageclaude-haiku-4-5-202510015/5Automated statement and correspondence generation costs pennies per item once infrastructure is in place, orders of magnitude cheaper than a teller's loaded hourly wage for composing, typing, and mailing individual letters.
Cost vs. human wageclaude-sonnet-54/5Automated letter generation and mail merge systems are far cheaper per item than a teller manually composing and typing each letter.
Technical feasibility todayclaude-haiku-4-5-202510014/5Document generation, mail merge, and basic correspondence composition are mature capabilities deployed in banking and financial services today. However, some edge cases (unusual discrepancies, sensitive customer issues) may require human oversight, preventing a perfect 5.
Technical feasibility todayclaude-sonnet-53/5Banking back-office systems increasingly use templated/automated correspondence generation, but full end-to-end composition tied to real-time account discrepancy detection and mailing is not universally deployed at tellers' level.

Examine checks for endorsements and to verify other information, such as dates, bank names, identification of the persons receiving payments, and the legality of the documents.

72

CI 70–74 · exposure 75 · augmentation 75 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Major financial institutions have already deployed automated check-processing and verification systems in production, and adoption has been steady in the banking sector—a highly digitized, capital-intensive industry with strong incentives to automate routine document inspection.
Sector adoption velocityclaude-sonnet-54/5Banking is a highly digitized sector with mobile deposit and automated check processing already deeply adopted, though teller-facing in-branch verification still involves humans for now.
Augmentation potentialclaude-haiku-4-5-202510014/5AI systems can flag suspicious or ambiguous endorsements and present structured check data to human tellers, significantly accelerating their inspection workflow and reducing manual transcription errors while keeping the human in a verification loop.
Augmentation potentialclaude-sonnet-54/5AI-assisted image capture, OCR, and fraud-flagging tools significantly speed up and improve accuracy of teller verification tasks while a human remains for final judgment and customer interaction.
Task automatabilityclaude-haiku-4-5-202510014/5Checks can be image-processed and their fields extracted with high accuracy by current OCR and document-understanding AI systems. Verification of endorsements, dates, bank names, and identification information are largely pattern-matching tasks that AI can perform end-to-end with significant time savings, though manual oversight may remain standard practice.
Task automatabilityclaude-sonnet-54/5Check verification (endorsement presence, date validity, bank name matching, basic fraud flags) is largely rule-based and pattern-matching, which OCR/computer vision and automated check-processing systems already handle at scale in remote deposit capture and ATM systems.
Adoption barriersclaude-haiku-4-5-202510013/5Banking is heavily regulated and many institutions maintain human oversight for fraud and legal compliance reasons; liability concerns and legacy internal controls create modest friction, but no strict legal requirement that a human personally verify every check prevents automation.
Adoption barriersclaude-sonnet-53/5No licensing requirement for check examination itself, but identity verification and fraud liability create moderate institutional caution and compliance requirements (KYC/BSA rules) that slow full removal of human oversight.
Cost vs. human wageclaude-haiku-4-5-202510015/5Check-processing AI inference is extremely cheap (pennies per document) and requires minimal integration overhead compared to the fully-loaded cost of a teller performing manual inspection, representing an order-of-magnitude cost advantage.
Cost vs. human wageclaude-sonnet-54/5Automated image-based check processing costs fractions of a cent per transaction versus a teller's loaded wage for manual review, though initial system integration adds cost.
Technical feasibility todayclaude-haiku-4-5-202510014/5Deployed products (document verification systems, check-processing automation in banking) already handle endorsement detection and field extraction in production; major banks use AI-driven check scanning and validation at scale, though often with human review embedded.
Technical feasibility todayclaude-sonnet-54/5Banks widely deploy automated check-image analysis and fraud-detection systems (e.g., ATM/mobile deposit capture) that verify these elements in production, though human tellers still handle edge cases and in-person identity verification.

Obtain and process information required for the provision of services, such as opening accounts, savings plans, and purchasing bonds.

71

CI 69–74 · exposure 75 · augmentation 75 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Financial services is a digitization leader with rapid AI adoption in the information sector. Major banks actively deploy chatbots, RPA, and automated account-opening systems; however, adoption is still not universal across all institution types and geographies, and regulatory caution slows some deployments.
Sector adoption velocityclaude-sonnet-54/5Retail banking has aggressively digitized account opening and self-service transactions over the past decade, a fast-adopting information-intensive sector.
Augmentation potentialclaude-haiku-4-5-202510014/5AI assistants substantially augment tellers by auto-populating forms, suggesting account types, flagging compliance issues, and retrieving customer history instantly. This dramatically accelerates service delivery while the teller remains responsible for final approval and customer interaction.
Augmentation potentialclaude-sonnet-54/5AI-assisted data entry, form pre-fill, and verification tools significantly speed up the work of remaining staff who handle exceptions or complex requests.
Task automatabilityclaude-haiku-4-5-202510014/5Most steps in account opening, savings plan setup, and bond purchasing involve rule-based data collection, form filling, KYC/AML checks, and document verification—all achievable by current AI systems at significant time savings. However, complex edge cases or unusual customer situations may still require human intervention, preventing a perfect 5.
Task automatabilityclaude-sonnet-54/5Data collection, identity verification, and account/bond setup workflows are highly structured and already handled by online banking platforms and chatbots with minimal human involvement in many cases.
Adoption barriersclaude-haiku-4-5-202510014/5Strong barriers exist: regulatory oversight (AML/KYC requirements), compliance sign-off, and liability risk if AI makes errors in account creation or fraud detection. Many jurisdictions require human review or authorization of sensitive account activities, and some customers prefer human verification for account security.
Adoption barriersclaude-sonnet-53/5KYC/AML regulations require identity verification and recordkeeping, and some customers still prefer human interaction, but these are compliance/process barriers rather than requirements for licensed human execution.
Cost vs. human wageclaude-haiku-4-5-202510015/5AI-driven account opening and data processing (via RPA, chatbots, and document automation) costs a fraction of a human teller's loaded wage. Once systems are deployed, per-transaction costs are negligible compared to even minimum wage labor for repetitive data entry and verification.
Cost vs. human wageclaude-sonnet-55/5Digital account-opening systems process thousands of applications at a fraction of the marginal cost of a teller performing the same paperwork-driven task.
Technical feasibility todayclaude-haiku-4-5-202510014/5Many banks have deployed AI and RPA systems that handle account opening workflows, document verification, and basic compliance checks in production. Mature solutions exist from fintech and legacy banking systems, though complete end-to-end automation without human oversight remains less common at the strictest compliance thresholds.
Technical feasibility todayclaude-sonnet-54/5Online account opening, savings plan enrollment, and bond purchases are already deployed at scale by banks and brokerages via web/app self-service and backend automation, though some edge cases still route to staff.

Answer telephones and assist customers with their questions.

67

CI 54–81 · exposure 62 · augmentation 75 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Financial services and banking have deeply integrated IVR and chatbot agents for inbound customer service; adoption is mature and widespread in large institutions.
Sector adoption velocityclaude-sonnet-53/5Financial services broadly adopt AI customer service tools, but retail banking branch/teller functions lag behind more digitized banking channels in full AI deployment.
Augmentation potentialclaude-haiku-4-5-202510014/5AI assistants can automatically retrieve customer data, suggest responses, and draft replies to incoming calls, significantly boosting teller productivity when they do engage directly with callers.
Augmentation potentialclaude-sonnet-54/5AI-assisted call routing, knowledge-base lookup, and real-time suggested responses can meaningfully speed up tellers' ability to answer customer questions accurately.
Task automatabilityclaude-haiku-4-5-202510014/5Current AI systems (chatbots, voice agents) can handle routine inquiries about account balances, transaction history, and basic troubleshooting with 50%+ time savings for a large portion of incoming calls, though complex or escalation scenarios still require human intervention.
Task automatabilityclaude-sonnet-53/5Voice AI and chatbots can handle routine account questions, balance inquiries, and FAQs, but complex or sensitive banking issues still require human handling for accuracy and trust, limiting full automation to roughly half the volume.
Adoption barriersclaude-haiku-4-5-202510013/5Banks deploy these systems routinely, but customer preference for human contact on sensitive matters, regulatory oversight of financial advice, and reputational risk create moderate friction against full automation.
Adoption barriersclaude-sonnet-52/5No strict licensing requirement for phone support, but banking regulations around identity verification, fraud prevention, and customer trust create moderate friction against full automation.
Cost vs. human wageclaude-haiku-4-5-202510015/5Per-call inference cost for voice agents is now a few cents, while a teller's loaded wage translates to roughly $15–25 per call; automation achieves 100x+ cost advantage for routine interactions.
Cost vs. human wageclaude-sonnet-54/5Automated phone/chat systems are far cheaper per interaction than a teller's loaded wage for routine queries, though complex cases still require costlier human intervention blended into the average.
Technical feasibility todayclaude-haiku-4-5-202510014/5Deployed IVR and conversational AI systems are widely used in banking today (e.g., major banks' automated customer service), successfully handling a substantial fraction of inbound calls and reducing human teller involvement for routine questions.
Technical feasibility todayclaude-sonnet-53/5Banks have deployed IVR systems and AI chat/voice assistants for basic inquiries, but these have material error rates and are often narrow in scope, escalating complex questions to humans.

Cash checks and pay out money after verifying that signatures are correct, that written and numerical amounts agree, and that accounts have sufficient funds.

66

CI 51–81 · exposure 62 · augmentation 63 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Banking and financial services have aggressively adopted automated check processing, ACH transfers, and self-service channels for decades. ATMs and digital banking have already displaced millions of teller roles; AI-enhanced verification is a natural continuation of this trend.
Sector adoption velocityclaude-sonnet-54/5Banking is a fast-digitizing sector with heavy investment in ATMs, ITMs, and mobile deposit, though branch teller roles persist for edge cases and relationship banking.
Augmentation potentialclaude-haiku-4-5-202510014/5AI assists tellers by automatically flagging mismatched amounts, suspicious signatures, and insufficient-fund warnings, dramatically reducing manual review time and error rates. The teller remains in control but works much faster with AI-generated alerts and recommendations.
Augmentation potentialclaude-sonnet-53/5AI-assisted fraud detection and signature/amount verification tools help tellers work faster and more accurately, though the core cash transaction still often requires human handling.
Task automatabilityclaude-haiku-4-5-202510014/5Modern AI systems can automate most of this task: signature verification via computer vision, amount reconciliation via OCR and numerical matching, and fund verification via database queries. However, edge cases (disputed signatures, fraud detection nuance) and occasional manual override reduce it from 5 to 4.
Task automatabilityclaude-sonnet-53/5Signature and amount verification plus balance checks are largely rule-based and already automated in ATMs and mobile deposit systems, but the physical cash handling and in-person judgment component resists full automation.'
Adoption barriersclaude-haiku-4-5-202510014/5Significant barriers exist: regulatory compliance (Know Your Customer, AML), fraud liability (banks bear loss if verification fails), and customer preference for human interaction on sensitive transactions. Many jurisdictions still require human sign-off on large or suspicious transactions.
Adoption barriersclaude-sonnet-52/5Some regulatory and fraud-liability considerations exist for large transactions, but automated check cashing and cash dispensing are already legally and operationally normalized via ATMs/ITMs.
Cost vs. human wageclaude-haiku-4-5-202510015/5Automated check processing and cash dispensing (ATMs, sortation systems) cost far less per transaction than a bank teller's fully-loaded wage (~$35–45k annually), especially at scale. AI-driven verification is a fraction of pennies per check.
Cost vs. human wageclaude-sonnet-54/5ATMs and automated check processing cost far less per transaction than a loaded teller wage, though cash-handling hardware and maintenance add some cost versus pure software tasks.
Technical feasibility todayclaude-haiku-4-5-202510013/5Check-processing automation exists in production (many banks use it), but end-to-end teller cash-handling including physical currency disbursement and real-time customer interaction remains partially manual. Products handle document verification well but not the full human interaction context.
Technical feasibility todayclaude-sonnet-54/5ATMs, ITMs (interactive teller machines), and mobile check deposit with OCR/fraud detection are mature, widely deployed products handling this exact workflow at scale.

Issue checks to bond owners in settlement of transactions.

66

CI 60–71 · exposure 66 · augmentation 75 · importance 3.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Banking and financial services sectors have rapidly adopted automated check systems, ACH, and settlement automation over the past decade; many institutions already deploy robotic process automation and API-based check issuance in production.
Sector adoption velocityclaude-sonnet-54/5Banking and financial services are among the fastest sectors adopting automation and straight-through processing for back-office transactions like settlements.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can significantly assist tellers by auto-populating check details, flagging suspicious transactions, verifying recipient data, and streamlining settlement workflows, allowing humans to focus on exceptions and compliance verification while maintaining control.
Augmentation potentialclaude-sonnet-54/5AI and automated systems significantly speed up verification, calculation, and check generation, letting tellers focus on exceptions and customer service while the routine processing is augmented.
Task automatabilityclaude-haiku-4-5-202510014/5The task of issuing checks involves deterministic processes (check printing, signing, recipient verification) that can be largely automated with current systems; however, settlement of bond transactions may require human verification of account status and regulatory compliance in some jurisdictions, preventing full end-to-end automation without oversight.
Task automatabilityclaude-sonnet-54/5Issuing checks for bond settlements is a rules-based, structured data-processing task involving verification and payment issuance, which existing automated banking/settlement systems can handle with high time savings once integrated.
Adoption barriersclaude-haiku-4-5-202510014/5Financial regulations (Know Your Customer, anti-fraud, settlement rules) and fiduciary duties often require human sign-off or authorization on bond settlement transactions; liability for incorrect issuance creates legal and compliance barriers that prevent pure automation in many institutions.
Adoption barriersclaude-sonnet-53/5Financial transactions involving bond settlements carry regulatory compliance, fraud liability, and audit requirements that create moderate friction, though not a strict licensed-human-signoff requirement for the mechanical check issuance itself.
Cost vs. human wageclaude-haiku-4-5-202510015/5Automated check issuance via banking infrastructure costs pennies per transaction (printing, routing, verification) compared to 15–30 minutes of teller labor at $20–30/hour loaded cost, making AI/automation at least 10× cheaper.
Cost vs. human wageclaude-sonnet-54/5Automated payment/settlement processing is dramatically cheaper per transaction than manual teller labor once systems are integrated, though initial setup and compliance costs offset some savings.
Technical feasibility todayclaude-haiku-4-5-202510013/5Banking software and payment systems can automate check generation and issuance, but real-world deployment typically requires human tellers to verify transaction legitimacy and manage edge cases; fully autonomous check issuance for bond settlements exists in limited contexts with significant manual oversight.
Technical feasibility todayclaude-sonnet-53/5Automated settlement and disbursement systems exist and are used in banking, but many teller-level check issuance workflows still involve manual verification steps and legacy systems, so reliability varies by institution.

Carry out special services for customers, such as ordering bank cards and checks.

58

CI 41–75 · exposure 62 · augmentation 63 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Large banks have piloted automated ordering systems, but adoption remains uneven; many smaller and regional banks still rely on manual teller-driven processes due to legacy systems and compliance complexity.
Sector adoption velocityclaude-sonnet-54/5Retail banking has aggressively digitized routine account services over the past decade, with online/mobile self-service now the default channel for such requests.
Augmentation potentialclaude-haiku-4-5-202510014/5AI-powered order systems significantly assist tellers by auto-populating customer data, validating requests, and tracking status, allowing them to process more orders faster while maintaining compliance checks.
Augmentation potentialclaude-sonnet-53/5AI-assisted systems help tellers or call-center staff quickly process these requests, but the augmentation value is modest since the task is largely already automated rather than augmented.
Task automatabilityclaude-haiku-4-5-202510013/5Ordering bank cards and checks involves structured data entry and form completion, which AI can handle; however, customer identity verification, account linking, and fulfillment coordination require human oversight, limiting end-to-end automation to roughly 50% of the task.
Task automatabilityclaude-sonnet-54/5Ordering bank cards or checks is a structured, rules-based transaction that chatbots, IVR systems, and online banking portals already handle with minimal human input. Most of the workflow (identity verification, form completion, order submission) can be automated end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Banks face regulatory requirements for identity verification, authorization, and audit trails; customer preference for human interaction on financial services and liability for errors create strong institutional friction against full automation.
Adoption barriersclaude-sonnet-52/5No licensing requirement for this specific task, but banks maintain some human backup for identity verification and error correction, creating mild organizational friction.
Cost vs. human wageclaude-haiku-4-5-202510012/5Automating card and check ordering requires infrastructure integration, compliance oversight, and error handling that offsets the low per-transaction cost; human tellers remain cheaper for low-volume branches.
Cost vs. human wageclaude-sonnet-54/5Automated self-service card/check ordering costs a fraction of a cent per transaction in server/API costs versus a teller's loaded wage for the same task, though integration and fraud-check overhead reduce the ratio slightly.
Technical feasibility todayclaude-haiku-4-5-202510013/5Some banking software includes automated ordering workflows, but systems still rely on human tellers to verify customer identity and process requests reliably; production deployment exists but with material variation in integration across institutions.
Technical feasibility todayclaude-sonnet-54/5Major banks already deploy online/mobile banking and chat-based systems that let customers order cards and checks without teller involvement, though some edge cases still route to a human.

Process and maintain records of customer loans.

58

CI 45–71 · exposure 62 · augmentation 75 · importance 3.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5The financial services sector, especially retail banking, has been among the fastest adopters of automation and AI; loan processing automation is already deeply embedded in production systems across most major institutions.
Sector adoption velocityclaude-sonnet-53/5Banking is a digitized industry with growing AI/RPA use for back-office tasks, but teller-level loan record processing still lags behind areas like customer service chatbots or fraud detection in adoption depth.
Augmentation potentialclaude-haiku-4-5-202510014/5AI assists tellers and loan officers by auto-populating forms, flagging compliance issues, summarizing customer history, and surfacing relevant loan terms, substantially raising human productivity while humans retain control over exceptions and final decisions.
Augmentation potentialclaude-sonnet-54/5AI-assisted data entry, OCR for documents, and automated validation significantly speed up record processing and reduce errors while tellers remain responsible for final verification and customer interaction.
Task automatabilityclaude-haiku-4-5-202510014/5Current AI and workflow automation systems can handle most of the task end-to-end: data entry, compliance verification, document processing, and record-keeping all have mature automation components. However, some discretionary judgment in loan assessment and exception handling requires human oversight, preventing a full 5 rating.
Task automatabilityclaude-sonnet-53/5Loan record processing involves structured data entry, document verification, and updates that AI/RPA can handle, but exceptions, compliance checks, and customer verification still require human judgment, limiting full automation to roughly half the workflow.
Adoption barriersclaude-haiku-4-5-202510014/5Banking and loan operations face strong regulatory oversight (compliance, audit trails, regulatory capital rules) and liability concerns that mandate documented audit paths and human sign-off on certain decisions, though routine processing itself is heavily automated.
Adoption barriersclaude-sonnet-54/5Financial record-keeping is heavily regulated (KYC, lending compliance, audit trails), often requiring human sign-off and accountability for accuracy and fraud prevention, creating substantial barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510015/5Automated loan processing and record maintenance cost a fraction of manual teller labor; once implemented, per-transaction costs are orders of magnitude lower than loaded human wages for equivalent output.
Cost vs. human wageclaude-sonnet-53/5Automation software and AI tools reduce processing time but require licensing, integration, and compliance oversight costs that keep the ratio closer to parity with teller wages rather than an order-of-magnitude cheaper.
Technical feasibility todayclaude-haiku-4-5-202510014/5Deployed banking automation systems, RPA platforms, and loan management software reliably process and maintain loan records in production at scale across major financial institutions. Minor edge cases and complex exceptions still require human review, keeping this below a perfect 5.
Technical feasibility todayclaude-sonnet-53/5Banking core systems and RPA tools already automate parts of loan record maintenance, but most deployments still require human tellers to input, verify, and reconcile records, so it's not fully hands-off in production.

Inform customers about foreign currency regulations and compute transaction fees for currency exchanges.

54

CI 46–62 · exposure 58 · augmentation 75 · importance 3.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Banking automation in back-office operations is fast, but customer-facing teller tasks—especially those involving regulatory advice—see slower adoption in production due to compliance risk aversion and customer preference for human interaction. Pilots are common, but widespread deployment remains limited.
Sector adoption velocityclaude-sonnet-53/5Banking is a high-digitization sector adopting AI broadly, but teller-facing FX services still see slower rollout of full self-service automation compared to core banking IT functions.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can meaningfully assist tellers by instantly retrieving applicable regulations, calculating fees accurately, and suggesting compliance checks, allowing the human to focus on customer service and judgment calls. This combination of reliable lookup + calculation with human oversight is already deployed in some banking systems.
Augmentation potentialclaude-sonnet-54/5AI tools can quickly surface current regulations and auto-calculate fees, letting tellers focus on customer interaction and exception handling, substantially speeding routine parts of the task.
Task automatabilityclaude-haiku-4-5-202510013/5AI can reliably compute transaction fees using rule-based engines and retrieve/summarize currency regulations, but must handle diverse, evolving regulatory frameworks across jurisdictions and occasional exceptions that require human judgment. This covers roughly half the task's value with significant upfront setup.
Task automatabilityclaude-sonnet-54/5Providing regulatory information and computing fees is largely rule-based lookup and arithmetic, which chatbots and calculators handle well; only edge-case regulatory nuance and in-person cash handling limit full automation.
Adoption barriersclaude-haiku-4-5-202510014/5Banking and currency exchange are heavily regulated; compliance and audit trails often require human sign-off or authorization, and liability for incorrect regulatory advice or fee miscalculation creates organizational and legal friction. Customers may also prefer human contact for trust in currency matters.
Adoption barriersclaude-sonnet-53/5Currency exchange involves compliance (AML/KYC) and disclosure requirements that push some institutions to require staff oversight, though this is not a hard licensing requirement everywhere.
Cost vs. human wageclaude-haiku-4-5-202510014/5AI inference for fee calculation and regulation lookup is very cheap per transaction (fractions of a cent), while a human teller's fully-loaded wage cost per customer interaction is orders of magnitude higher. The cost advantage is substantial, though integration and compliance oversight add overhead.
Cost vs. human wageclaude-sonnet-54/5Automated fee calculators and rate lookups cost fractions of a cent per transaction versus a teller's loaded wage, though initial integration and compliance monitoring add some cost.
Technical feasibility todayclaude-haiku-4-5-202510013/5Current AI systems can compute standard fees and provide general information about currency regulations via lookup and summarization, but production deployments in banking remain limited by compliance sensitivities and the need for human verification of regulatory interpretations. Demos exist but real-world banking adoption remains cautious.
Technical feasibility todayclaude-sonnet-53/5Bank chatbots and ATMs/kiosks already provide FX rates and fee calculations, but full conversational regulatory guidance with accuracy guarantees is less mature in production teller-replacement systems.

Count, verify, and post armored car deposits.

52

CI 30–75 · exposure 50 · augmentation 50 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Banks and armored car services have already widely deployed automated counting and verification systems in production; deposit automation is well-established in the financial services sector where digitization is high.
Sector adoption velocityclaude-sonnet-52/5Banking back-office cash operations have seen some automation (cash recyclers, deposit automation), but this specific physical task of armored car deposit handling remains largely manual and adoption is slow compared to digital-only banking tasks.:
Augmentation potentialclaude-haiku-4-5-202510013/5Automated counting and verification systems assist tellers by handling routine verification and flagging discrepancies, freeing them for exception handling and customer interaction, but the augmentation is bounded by the task's primarily mechanical nature.
Augmentation potentialclaude-sonnet-53/5Software tools can assist tellers with verification, reconciliation, and flagging discrepancies in deposit records, improving speed and accuracy while the human still performs physical counting and oversight.:
Task automatabilityclaude-haiku-4-5-202510014/5Counting and verifying deposits can be largely automated using optical scanning, image recognition, and machine learning to validate denominations and totals. However, posting to core banking systems requires integration with legacy systems and manual oversight for reconciliation discrepancies, preventing full end-to-end automation from reliably achieving 50%+ time savings without human intervention.
Task automatabilityclaude-sonnet-52/5Physical handling of armored car deposits requires manual counting and verification of physical cash/checks, which current AI cannot perform end-to-end without robotic hardware; only the posting/reconciliation portion is automatable today.:
Adoption barriersclaude-haiku-4-5-202510012/5There are few legal or regulatory barriers to automating deposit counting and verification; banks can deploy these systems independently. Organizational friction around legacy system integration and staff retraining present minor friction but do not block adoption.
Adoption barriersclaude-sonnet-53/5Banking regulations require accurate record-keeping and audit trails for cash handling, and dual-control/verification procedures often mandate human oversight, creating moderate friction against full automation.:
Cost vs. human wageclaude-haiku-4-5-202510014/5Automated currency counting and verification systems cost a fraction of a teller's loaded wage per transaction, with minimal marginal cost per deposit processed. Integration and maintenance overhead is modest relative to labor savings.
Cost vs. human wageclaude-sonnet-52/5Automated cash-counting and deposit-processing equipment has meaningful capital and maintenance costs comparable to teller labor, so while some efficiency gains exist, it is not dramatically cheaper than human labor for this specific task.:
Technical feasibility todayclaude-haiku-4-5-202510014/5Deployed currency counting machines with image recognition and validation systems are in production use at banks today; armored car services use automated counting and verification. However, the full posting workflow including system integration and exception handling still requires human oversight, limiting reliability to 4 rather than 5.
Technical feasibility todayclaude-sonnet-52/5Cash-counting machines and automated deposit reconciliation software exist and are used in production, but the full task including physical verification and exception handling still relies heavily on human tellers or specialized cash-processing equipment, not general AI systems.:

Prepare and verify cashier's checks.

47

CI 25–70 · exposure 50 · augmentation 63 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Banking has moderate digital adoption, but cashier's check preparation is a small, declining service within teller operations as digital payment methods grow, and banks remain conservative about automating high-liability financial instruments without human oversight.
Sector adoption velocityclaude-sonnet-54/5Retail banking has rapidly adopted automated teller systems, online banking, and backend automation for check processing, though full teller replacement varies by institution size.
Augmentation potentialclaude-haiku-4-5-202510013/5AI could assist by auto-populating check fields from customer data, flagging suspicious requests, or pre-verifying standard compliance checks, meaningfully speeding up the teller's workflow while they retain control over final authorization and verification.
Augmentation potentialclaude-sonnet-54/5AI and automated systems significantly speed up verification, fraud detection, and document preparation, allowing tellers to process these tasks faster while retaining oversight.
Task automatabilityclaude-haiku-4-5-202510012/5Bank tellers prepare checks by filling in amounts, dates, and recipient information, which modern AI could partially automate via document generation. However, the verification step requires reconciliation against accounts, fraud detection, and compliance checks that demand human judgment and access to internal systems, limiting end-to-end automation to a small portion of the full task.
Task automatabilityclaude-sonnet-54/5Preparing and verifying cashier's checks is a structured, rules-based process (data entry, balance verification, fee calculation) that core banking systems and workflow automation can largely handle end-to-end today.》Most steps are already digitized within bank software.》
Adoption barriersclaude-haiku-4-5-202510014/5Banking regulations (including BSA/AML and check certification requirements) legally mandate human involvement in cashier's check issuance and verification. Liability concerns are severe—errors on cashier's checks are the bank's responsibility—creating strong regulatory and legal barriers to automation.
Adoption barriersclaude-sonnet-53/5Banking regulations require certain verification and fraud controls, and some institutions mandate human sign-off for official checks, creating moderate compliance friction.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI inference for document preparation is cheap, but integration with banking systems, compliance verification, and the overhead of human oversight for high-liability financial instruments makes the all-in cost comparable to or potentially higher than a teller's partial hourly wage.
Cost vs. human wageclaude-sonnet-54/5Automated check issuance via software is far cheaper per transaction than paying a teller's loaded wage, though some human oversight and system costs remain.
Technical feasibility todayclaude-haiku-4-5-202510012/5While form-filling and basic document generation are mature, no deployed product reliably handles the full task of preparing and verifying cashier's checks in production banking systems, which require integration with core banking infrastructure, fraud detection, and regulatory compliance.
Technical feasibility todayclaude-sonnet-54/5Core banking platforms and automated teller/branch systems already generate and verify cashier's checks with minimal human intervention in many banks, though some manual review persists for fraud checks.

Explain, promote, or sell products or services, such as travelers' checks, savings bonds, money orders, and cashier's checks, using computerized information about customers to tailor recommendations.

34

CI 23–45 · exposure 33 · augmentation 75 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Banks have moved toward digital self-service and ATMs for routine transactions, but consultative selling of complex financial products remains a human-staffed function. Adoption of AI-driven sales in banking is slow and limited to narrow use cases like chatbot initial screening.
Sector adoption velocityclaude-sonnet-53/5Banking is adopting AI for personalization and recommendation engines at a moderate pace, though teller-facing cross-sell tools are less mature than back-office automation.
Augmentation potentialclaude-haiku-4-5-202510014/5AI systems can significantly assist tellers by surfacing relevant customer data, suggesting products based on profile, and handling data lookup in real time. This augmentation boosts teller productivity and cross-sell rates while the human teller retains relationship and compliance responsibility.
Augmentation potentialclaude-sonnet-54/5AI-based customer analytics can meaningfully assist tellers by surfacing relevant product suggestions and next-best-action prompts, improving the effectiveness of their conversations.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can retrieve customer data and generate product recommendations, the task requires persuasion and relationship-building that depends on reading social cues, objection handling, and contextual judgment. Current AI systems can support recommendation generation but cannot reliably handle the full persuasion loop end-to-end with quality equal to human tellers.
Task automatabilityclaude-sonnet-53/5AI can generate personalized product recommendations and scripts from customer data, but the actual customer interaction, trust-building, and sale still typically require a human teller or at least a conversational agent with limited deployment in this exact context.
Adoption barriersclaude-haiku-4-5-202510014/5Banking is heavily regulated; sales of financial products like savings bonds and cashier's checks often require human authorization, licensing (Series 6/7 in the US), and personal accountability. Customer authentication and suitability determinations carry legal liability that restricts autonomous AI deployment.
Adoption barriersclaude-sonnet-53/5No licensing requirement for basic teller sales, but banks have compliance, disclosure, and consumer-protection expectations that favor human involvement in financial product recommendations.
Cost vs. human wageclaude-haiku-4-5-202510012/5Implementing a compliant AI system for financial product sales, including oversight, integration with banking systems, and liability management, is expensive. The cost per transaction is likely comparable to or higher than paying a teller, especially when regulatory and error costs are included.
Cost vs. human wageclaude-sonnet-53/5AI-driven recommendation engines are cheap to run, but integrating them into teller workflows plus needed human oversight for compliance keeps costs roughly comparable to labor cost for this specific task.
Technical feasibility todayclaude-haiku-4-5-202510012/5Chatbots and recommendation engines exist for product suggestions, but no deployed product reliably performs consultative selling with financial products at the quality and compliance standards required by banking. Human tellers remain the standard for this task in production banking environments.
Technical feasibility todayclaude-sonnet-52/5Some banks use chatbots and CRM-driven recommendation engines, but cross-selling financial products via a teller interaction is not yet reliably automated end-to-end in production at scale.

Resolve problems or discrepancies concerning customers' accounts.

31

CI 25–36 · exposure 25 · augmentation 63 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While banking is digitized, autonomous account resolution automation remains limited in production; most adoption is in diagnostic and support tools rather than end-to-end problem resolution, reflecting regulatory caution and risk aversion.
Sector adoption velocityclaude-sonnet-53/5Financial services are moderate-to-fast adopters of AI, with chatbots and automated dispute systems in production, but full resolution of nuanced account issues remains largely human-driven.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist tellers by quickly identifying discrepancies, suggesting common resolutions, and retrieving relevant account history, meaningfully raising teller efficiency on routine cases while the teller retains decision-making authority.
Augmentation potentialclaude-sonnet-54/5AI tools can surface transaction histories, flag anomalies, and draft resolution steps, meaningfully speeding up a teller's diagnostic and resolution process while the human retains decision authority.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can flag discrepancies and suggest resolutions through data analysis, resolving account issues typically requires judgment about customer circumstances, authorization decisions, and occasional manual adjustments that fall short of the 50% time-saving threshold for end-to-end automation.
Task automatabilityclaude-sonnet-52/5Resolving account discrepancies often requires judgment, exception handling, and empathetic customer interaction that current AI cannot fully replicate end-to-end, though simple cases (e.g., balance mismatches) can be flagged or partially resolved by systems.
Adoption barriersclaude-haiku-4-5-202510014/5Banking regulation, fiduciary duty, and customer protection laws typically require a licensed employee or authorized representative to authorize account adjustments and sign off on dispute resolutions, creating strong legal barriers to full automation.
Adoption barriersclaude-sonnet-53/5Banking regulations, fraud liability, and customer authentication requirements create meaningful friction, though not an absolute licensing requirement like some financial advisory roles.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI infrastructure and integration costs are comparable to or exceed the cost of a teller handling routine discrepancy resolution, especially when considering oversight, compliance review, and error correction overhead.
Cost vs. human wageclaude-sonnet-53/5AI-assisted triage tools are cheaper for simple cases, but escalation to human review for complex discrepancies keeps overall cost comparable to human-only handling once integration and oversight are factored in.
Technical feasibility todayclaude-haiku-4-5-202510012/5Some banks deploy limited AI for routine discrepancy detection and categorization, but production systems do not reliably resolve complex or novel account problems without human intervention; most deployed solutions are diagnostic aids rather than autonomous resolvers.
Technical feasibility todayclaude-sonnet-52/5Some banks deploy chatbots and rule-based systems for basic dispute triage, but complex discrepancy resolution still routes to human tellers or specialists due to reliability and trust concerns.

Monitor bank vaults to ensure cash balances are correct.

25

CI 25–25 · exposure 25 · augmentation 50 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Banking has adopted AI for fraud detection and transaction monitoring, but vault monitoring remains largely manual and procedural. Adoption of AI-assisted vault monitoring is still pilot-stage; most institutions rely on existing human audit and reconciliation processes with slow organizational change in this risk-sensitive domain.
Sector adoption velocityclaude-sonnet-52/5Banking is adopting AI for fraud detection and customer service faster than for physical cash vault operations, which remain reliant on secure physical processes and human verification.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist tellers by automatically reconciling transaction records against recorded balances and alerting them to discrepancies, reducing manual counting time and improving accuracy. However, the core task—physical verification and accountability—still requires human presence and judgment, making augmentation partial rather than transformative.
Augmentation potentialclaude-sonnet-53/5Software systems can flag discrepancies and automate reconciliation calculations, helping tellers verify balances faster, though the physical monitoring and custody remain human-led.
Task automatabilityclaude-haiku-4-5-202510012/5Current AI can assist with detecting discrepancies in recorded balances and flagging anomalies, but physical vault monitoring requires real-time visual inspection, access control verification, and human judgment about security protocols that AI cannot fully replace. End-to-end automation would require autonomous physical inspection systems, which are not deployed in banking contexts today.
Task automatabilityclaude-sonnet-52/5Physical vault monitoring and cash reconciliation require presence and handling of physical currency, which AI cannot directly perform; only the counting/reconciliation math could be automated with hardware like cash recyclers, not general AI.
Adoption barriersclaude-haiku-4-5-202510014/5Banks face strong regulatory requirements (BSA, vault security protocols) and legal liability for cash accountability that effectively require human certification and physical presence. Compliance frameworks and insurance underwriting typically mandate human teller sign-off on vault balances, creating hard operational barriers to full automation.
Adoption barriersclaude-sonnet-54/5Cash custody involves strict internal controls, dual-custody requirements, and regulatory/audit obligations that typically mandate human accountability for vault balances.
Cost vs. human wageclaude-haiku-4-5-202510012/5Implementing AI systems for vault monitoring (cameras, sensors, analytics infrastructure) and the required human oversight still exceeds the cost of a teller's time for routine balance checks, especially given the low volume and infrequency of vault access in most branches.
Cost vs. human wageclaude-sonnet-52/5Specialized cash-counting and vault-tracking hardware/software requires significant capital investment and integration, often comparable to or exceeding teller labor costs for this narrow task.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI-powered analytics can reconcile ledgers and highlight balance discrepancies from transaction data, no deployed system independently monitors physical vaults or certifies cash correctness without human verification. Regulatory requirements and liability concerns mean humans remain legally accountable for vault integrity, limiting production AI autonomy.
Technical feasibility todayclaude-sonnet-52/5Automated cash-handling machines and vault management systems exist in some banks, but full AI-driven vault monitoring without human oversight is not standard practice today.

Order a supply of cash to meet daily needs.

25

CI 25–25 · exposure 25 · augmentation 50 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While financial institutions are digitizing operations, cash ordering remains embedded in compliance-heavy, human-authorized workflows. Adoption of pure AI automation is slow; most banks use forecasting tools but retain human decision-makers, reflecting institutional conservatism in cash operations.
Sector adoption velocityclaude-sonnet-52/5Retail banking back-office cash operations are moderately digitized but move slowly due to compliance and legacy system constraints, with automation focused on forecasting rather than full task replacement.
Augmentation potentialclaude-haiku-4-5-202510013/5AI-driven forecasting and dashboards can assist tellers by predicting cash needs and recommending order quantities, reducing manual calculation and analysis time. However, the task's inherent need for human judgment and authorization limits the transformative potential of augmentation.
Augmentation potentialclaude-sonnet-53/5AI-based demand forecasting tools can meaningfully assist tellers/branch managers in estimating cash needs, improving accuracy and efficiency while humans still execute and authorize orders.
Task automatabilityclaude-haiku-4-5-202510012/5AI could help forecast cash needs based on historical transaction data and prepare orders, but human judgment is required to account for unusual circumstances (holidays, events, economic conditions) and final authorization. Current systems cannot reliably handle the full decision-to-order workflow autonomously.
Task automatabilityclaude-sonnet-52/5This is a semi-structured operational task involving forecasting cash needs and coordinating with vault/branch systems, which requires physical and organizational coordination that current AI cannot fully execute end-to-end.},
Adoption barriersclaude-haiku-4-5-202510014/5Banks face strong regulatory oversight (anti-money laundering, cash controls, reporting requirements) and internal controls that mandate human authorization and record-keeping for cash orders. Liability for cash shortages creates audit and accountability requirements that cannot be delegated to AI without regulatory approval.
Adoption barriersclaude-sonnet-54/5Cash handling and ordering in banks is subject to internal controls, dual-authorization requirements, and regulatory/audit oversight, creating significant procedural barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5The cost of implementing and maintaining AI-based forecasting systems, plus required human oversight and compliance infrastructure, is comparable to or exceeds the cost of a teller spending 30 minutes daily on cash ordering. Significant integration costs apply.
Cost vs. human wageclaude-sonnet-52/5While forecasting algorithms are cheap, the task also requires human authorization, verification, and interaction with cash logistics providers, keeping all-in automation costs comparable to human labor.
Technical feasibility todayclaude-haiku-4-5-202510012/5While predictive analytics tools exist to estimate cash demand, no deployed bank product reliably automates the complete cash ordering task end-to-end. Banks still require human tellers or vault managers to review forecasts and authorize orders with legal accountability.
Technical feasibility todayclaude-sonnet-52/5Some cash forecasting software exists to recommend order quantities, but the actual ordering, authorization, and physical logistics remain human-driven in production banking environments.

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.