Banking Use-Case Mapper

Risk & Compliance High risk

AML Alert Review

Transaction and sanctions alerts arrive in volumes far beyond review capacity. Analysts assemble evidence across core alerting, customer, and transaction systems before dispositioning each one, and the great majority close as false positives.

Pattern mapping

Investigation / Case ManagementKnowledge & Research

GenAI opportunity

Drafts the disposition narrative — why this alert looks like a false positive — citing the specific transactions and customer facts it relied on, so the investigator reviews reasoning rather than reassembling it.

Grounded on

Customer KYC profileTransaction historyPrior alert dispositionsTypology guidance

Stays deterministic Alert scoring thresholds and case state transitions.

Human control

Mandatory investigator review before disposition; suspicious activity filings always human-signed

Risk & Compliance High risk

Negative News Screening

Thousands of news articles are reviewed daily for adverse themes. Reviewers judge materiality, extract the parties named, and establish whether any of them hold a relationship with the bank. Every step must be logged for the record.

Pattern mapping

Knowledge & ResearchInvestigation / Case Management

GenAI opportunity

Reads each article, judges materiality against defined adverse themes, and extracts the parties named — the judgment call keyword rules keep getting wrong in both directions.

Grounded on

Article textAdverse-theme definitionsClient and employee lists

Stays deterministic Name matching against customer lists, and the investigation audit log.

Human control

Compliance specialist reviews all confirmed party matches before remediation

Risk & Compliance High risk

Account Takeover Investigation

Takeover investigations span a ticketing system, a fraud-detection platform, and an external identity provider. The fragmentation produces slow response, inconsistent case documentation, and inconsistent decisions between analysts.

Pattern mapping

Investigation / Case ManagementKnowledge & Research

GenAI opportunity

Assembles a case narrative from activity scattered across ticketing, fraud, and identity systems, and proposes a recommendation with the evidence attached.

Grounded on

Account activity logsIdentity provider profilePrior takeover casesFraud playbooks

Stays deterministic Credential reset, ticket creation, and any account action.

Human control

Fraud analyst makes the confirmation decision and authorises credential reissue

Risk & Compliance High risk

Ultimate Beneficial Owner Matching

High-risk client profiles must be reconciled monthly against an official government corporate registry, with any discrepancy reported inside a fixed regulatory window. Company names rarely match exactly between the two sources.

Pattern mapping

Entity Resolution & MatchingData Investigation

GenAI opportunity

Handles the semantic tail — entities whose registry name and client-master name differ by abbreviation, legal suffix, or transliteration and will never match on string comparison.

Grounded on

Registry extractClient master recordNaming convention rules

Stays deterministic The exact-match pass first; only unmatched records reach the model.

Human control

Relationship manager confirms discrepancies before regulatory reporting

Risk & Compliance Critical risk

Counterparty Credit Risk Assessment

Credit assessments for clients in high-risk jurisdictions require aggregating disparate data, weighing qualitative alongside quantitative factors, and producing defensible documentation under compressed timelines. Second and third-order impacts of a macro event are especially hard to anticipate.

Pattern mapping

Data InvestigationKnowledge & ResearchControlled Content GenerationAgentic Workflow

GenAI opportunity

Interprets a macro event and proposes which stress scenarios to run, including the second and third-order impacts an analyst would take days to reason through.

Grounded on

News feedInternal exposure dataPrior scenario libraryRisk policy

Stays deterministic Risk engine execution and scenario parameter application.

Human control

Risk manager accepts or reinitiates the scenario set; risk committee owns the decision

Regulatory Reporting Critical risk

Regulatory Variance Analysis

Analysts manually investigate period-on-period movements in reported metrics using data, metadata, and prior explanations, then write the commentary that accompanies the submission.

Pattern mapping

Data InvestigationKnowledge & ResearchControlled Content Generation

GenAI opportunity

Explains what drove the movement in reviewable prose, grounded in the query results and the wording of prior period commentary.

Grounded on

Reporting data martsMetric definitions and lineagePrior period commentary

Stays deterministic Figure calculation and reconciliation to source. The model explains numbers; it never produces them.

Human control

Mandatory review and sign-off before any reporting use

Consumer Banking High risk

Mortgage Origination Review

Fragmented intake, manual document validation, and policy exceptions slow origination. Operations teams bottleneck on document review, debt-to-income and loan-to-value checks, and keeping the loan origination system current.

Pattern mapping

Document IntelligenceInvestigation / Case Management

GenAI opportunity

Reasons about policy exceptions — cases where the documents don't cleanly satisfy a rule and an underwriter would have to interpret intent.

Grounded on

Underwriting policyBorrower documentsProduct guidelines

Stays deterministic Debt-to-income and loan-to-value arithmetic, and loan origination system updates.

Human control

Exceptions escalate to a human underwriter; loan officer performs final review

Consumer Banking High risk

Credit Card Dispute Handling

Disputes move through intake, categorisation, merchant evidence request, evaluation, decision, and customer notification. Each handoff is manual, categorisation depends on individual judgment, and outcomes vary between handlers.

Pattern mapping

Document IntelligenceInvestigation / Case Management

GenAI opportunity

Classifies the dispute reason from the customer's free-text description and drafts the merchant evidence request in network-compliant language.

Grounded on

Dispute reason codesCard network rulesTransaction and account record

Stays deterministic Chargeback filing and regulatory SLA timers.

Human control

Human reviews before the dispute is filed with the payment network

Consumer Banking High risk

International Account Opening

Onboarding clients across jurisdictions is slowed by multilingual correspondence, acceptance policies that vary by country, inconsistent document formats, and manual creation of client records.

Pattern mapping

Document IntelligenceAgentic WorkflowControlled Content Generation

GenAI opportunity

Interprets documents across languages and drafts correspondence in the applicant's language, including the reasoning behind a country-based decline.

Grounded on

Country acceptance policySubmitted documentsApproved letter templates

Stays deterministic Sanctions screening, case creation, and relationship manager assignment rules.

Human control

Relationship manager reviews the assembled file; sanctions hits stop the workflow

Consumer Banking Moderate risk

Insurance Letter of Guarantee (Auto Lending)

Staff verify insurance claims by hand, work through email threads, draft letters of guarantee, and coordinate by phone. Manual entry produces mismatched vehicle identifiers and payoff amounts, which trigger lien-release denials and rework.

Pattern mapping

Document IntelligenceControlled Content GenerationAgentic Workflow

GenAI opportunity

Drafts the letter with the correct lien terms assembled from claim correspondence, the loan record, and the title system.

Grounded on

Loan recordInsurer claim correspondenceTitle recordLetter template

Stays deterministic Vehicle identifier and payoff amount matching, and lien release posting.

Human control

Compliance officer approves the letter before issue

Consumer Banking High risk

Third-Party Demand Processing

Legal demands against customer accounts arrive at high volume under tight service-level commitments. Each must be validated for legality, matched to the right accounts, and actioned in the system of record.

Pattern mapping

Document IntelligenceAgentic Workflow

GenAI opportunity

Reads the demand, types it, and reasons about legality and scope against jurisdiction-specific procedures.

Grounded on

Demand letterJurisdiction SOPsAccount records

Stays deterministic Account holds, system-of-record updates, and the SLA clock.

Human control

Compliance officer validates demand legality before any account action

Consumer Banking Low risk

Contact Centre Mail Intent Classification

Inbound mail and email are typed manually before any work starts. Quality control relies on sampling after the fact rather than prevention, and external events produce volume swings that break service levels.

Pattern mapping

Classification & Routing

GenAI opportunity

Detects intent from unstructured correspondence where customers describe the same request in wildly different words.

Grounded on

Intent taxonomyHistorical labelled examples

Stays deterministic Queue routing and case creation. Classification directs work; it decides nothing.

Human control

Unrecognised intents raise an exception to a human; no decisioning beyond routing

Consumer Banking Critical risk

Consumer Authentication from Correspondence

Agents authenticate consumers by searching records on national identifiers, validating attached documents against business rules by hand, and comparing personal data against credit files. Error risk is high and volume spikes strain service levels.

Pattern mapping

Document IntelligenceEntity Resolution & Matching

GenAI opportunity

Reads varied identity documents and reconciles the fields against the consumer record where formats and layouts differ.

Grounded on

Document imagesCredit file dataConsumer record

Stays deterministic Identifier lookup and the pass/fail authentication rule. The rule stays a rule.

Human control

Failed or ambiguous authentication escalates to a human agent

Commercial Banking High risk

Commercial Loan QC and QA

Loan closing and quality control depend on individual judgment across multiple handoffs, with reviewers context-switching between disconnected systems to verify more than fifty data points and chase missing files.

Pattern mapping

Document IntelligenceControlled Content Generation

GenAI opportunity

Compares the same data point as it appears across several documents and describes the discrepancy in terms a reviewer can act on.

Grounded on

Closing packet documentsQC checklist and policyPrior exception log

Stays deterministic Document presence checks and notification dispatch.

Human control

Lending operations reviews the exception summary and owns remediation

Commercial Banking High risk

Commercial Credit Analyst Support

Pre-underwriting requires extracting financials from business tax returns and entering them into the origination system, then researching the applicant's business, industry, and legal history from several sources.

Pattern mapping

Document IntelligenceKnowledge & Research

GenAI opportunity

Synthesises business, industry, and legal research into a due-diligence note, which is the genuinely slow half of pre-underwriting.

Grounded on

Legal research serviceIndustry reportsApplicant filings

Stays deterministic Tax form field extraction into the origination system and financial spreading calculations.

Human control

Credit analyst reviews and approves before the package reaches underwriting

Wealth / Capital Markets Moderate risk

Investment Research Pack

Analysts gather data from several market data sources, summarise it, research the company, and compile the result into a defined format. The work drives long hours and contributes to analyst burnout.

Pattern mapping

Knowledge & ResearchControlled Content Generation

GenAI opportunity

Summarises filings and news into the house format — the step that turns gathered data into something an analyst can hand over.

Grounded on

Company filingsMarket data platformsNews sourcesPack template

Stays deterministic Data retrieval and final document assembly.

Human control

Analyst validates the pack before any client or internal use

Wealth / Capital Markets High risk

Fund Accounting Reconciliation

End-of-day net asset value checks involve verifying position and cash balances, running control checks, and analysing reconciliation breaks. Decomposing an exception code and identifying the transactions behind it remains manual analysis.

Pattern mapping

Data InvestigationInvestigation / Case Management

GenAI opportunity

Decomposes an exception code into a plain explanation and names the likely transactions behind it, ruling out timing differences first.

Grounded on

Break detailReference dataPrior break resolutions

Stays deterministic Reconciliation matching itself and any journal posting.

Human control

Human review of reconciliation breaks before book or edit of transactions

Payments Moderate risk

Corporate Payments Onboarding

After sale, customers are set up in core payment systems manually. Data is extracted from standard documents and re-keyed, with the process running largely on spreadsheets and institutional memory.

Pattern mapping

Document IntelligenceAgentic Workflow

GenAI opportunity

Reads non-standard implementation documents into the fields the setup actually needs, replacing spreadsheet-and-memory transcription.

Grounded on

Onboarding documentsProduct configuration rules

Stays deterministic The system setup sequence and time-delayed actions. This process is mostly orchestration, not reasoning.

Human control

Onboarding specialist validates before accounts go live; invalid data alerts a human

Payments Moderate risk

Treasury Client Servicing

Client requests arrive by email in varied forms, each needing manual research across systems. Response times are slow, quality is uneven, and there is no view of where the delays originate.

Pattern mapping

Classification & RoutingInvestigation / Case ManagementControlled Content Generation

GenAI opportunity

Interprets requests that arrive in no fixed form and drafts a professional reply grounded in account data and product documentation.

Grounded on

Request correspondenceProduct documentationAccount dataReply templates

Stays deterministic Case updates and sending. A human approves every outbound client message.

Human control

Subject-matter expert reviews before any client response is sent

Payments High risk

Cash Application and Remittance Matching

High volumes of incoming payments arrive with remittance advice referencing multiple invoices in inconsistent formats. Staff reconcile each transaction by hand, and mis-keyed entries produce incorrect postings and misapplied cash.

Pattern mapping

Entity Resolution & MatchingDocument Intelligence

GenAI opportunity

Interprets free-text remittance advice referencing several invoices in inconsistent formats, and proposes matches with confidence attached.

Grounded on

Remittance documentsOpen invoice ledgerCustomer payment history

Stays deterministic Exact matching first, and all journal entry posting.

Human control

Human approves the matching file before journal entries are posted

Insurance Moderate risk

Claims Intake and First Notice of Loss

Claims arrive as unstructured packets through mobile, web, and email and are processed manually. Manual intake raises processing cost, slows settlement, and creates compliance exposure.

Pattern mapping

Classification & RoutingDocument Intelligence

GenAI opportunity

Reads the packet — images, notes, forms — and classifies loss type and severity well enough to route it correctly on arrival.

Grounded on

Claim submissionPolicy recordLoss type taxonomy

Stays deterministic Claim number issue and acknowledgement letter dispatch.

Human control

Subject-matter expert review before claim creation; low-confidence packets held

Insurance High risk

Health Spending Account Claims

Participants submit reimbursement requests with receipts through app or portal. Each is checked for eligibility against tax-authority rules, reviewed, approved or denied, paid, and logged for audit.

Pattern mapping

Document IntelligenceAgentic Workflow

GenAI opportunity

Judges whether an expense qualifies under published rules in the cases where the receipt is ambiguous or the expense sits near a boundary.

Grounded on

Receipt imagesQualified expense rulesPlan terms

Stays deterministic Payment issue and audit recordkeeping.

Human control

Claims adjuster reviews flagged and exception claims before payment

Insurance High risk

Accident Claims Adjudication

Fragmented enrolment data makes eligibility hard to confirm, incomplete documentation slows intake, and judging whether an incident qualifies as accidental is often unclear. Complex policy terms require careful manual reading.

Pattern mapping

Document IntelligenceInvestigation / Case ManagementControlled Content Generation

GenAI opportunity

Extracts claimed services from medical documentation and reasons about whether the incident meets the policy's definition of accidental.

Grounded on

Medical bills and clinical notesGroup plan termsEnrolment data

Stays deterministic Benefits explanation generation and payment initiation.

Human control

Reviewer validates extraction and coverage match before benefits explanation is issued

Insurance High risk

Home Policy Eligibility Determination

Underwriters reconcile prior loss reports, inspection data, and applicant documents against policy rules by hand. The effort slows issuance, introduces inconsistency, and caps throughput at high application volumes.

Pattern mapping

Document IntelligenceInvestigation / Case ManagementAgentic Workflow

GenAI opportunity

Reconciles inconsistent prior-loss and inspection data and articulates why an applicant does or doesn't qualify — the audit trail underwriters currently can't produce.

Grounded on

Prior loss reportsProperty inspection dataEligibility rules

Stays deterministic Rule evaluation and quote generation.

Human control

Exceptions route to a human underwriter; declines require human confirmation

Insurance Moderate risk

Endorsement Materiality Screening

Requests to amend existing policies arrive daily. A team reads each email and its attachments to judge whether the requested change is material or purely administrative, then routes it to the matching queue.

Pattern mapping

Classification & RoutingKnowledge & Research

GenAI opportunity

Judges material against purely administrative from how the request is worded, which is exactly the line a keyword rule cannot hold.

Grounded on

Request correspondence and documentsMateriality definitionsPolicy terms

Stays deterministic Queue routing. Nothing about the policy changes at this step.

Human control

Ambiguous cases route to an underwriter; classification does not alter the policy

Insurance Moderate risk

Broker Submission Handling

Submissions arrive by email, producing inconsistency and compliance exposure. Duplicate and incomplete submissions consume underwriter time, and slow responses to brokers cost business.

Pattern mapping

Document IntelligenceAgentic Workflow

GenAI opportunity

Assesses appetite fit against underwriting guidelines and drafts the broker reply, which is where the competitive turnaround time is won.

Grounded on

Submission documentsAppetite guidelinesPrior submissions for duplicate detection

Stays deterministic Premium calculation, CRM updates, and folder structure creation.

Human control

Underwriter performs final review and owns all exceptions and high-value decisions

Insurance Moderate risk

Quote Negotiation Data Extraction

Thousands of emails pass between brokers and carrier underwriters each month to request and negotiate quotes. Official quotes sit inside PDFs or email bodies, leaving negotiation history unavailable for market analysis.

Pattern mapping

Document IntelligenceData Investigation

GenAI opportunity

Identifies negotiation stage and intent from email prose — the signal that exists nowhere in structured form and is the whole point of the exercise.

Grounded on

Email threadQuote documentsTarget field schema

Stays deterministic Field validation, storage, and dashboard load.

Human control

Broker validates extracted fields where confidence is low

Insurance Moderate risk

Benefits Booklet Summarisation

Carriers send benefits booklets tailored to different employee classes. Analysts read each one, identify covered benefits by class, and populate a standardised comparison template by hand.

Pattern mapping

Document IntelligenceControlled Content Generation

GenAI opportunity

Locates covered benefits by employee class inside long unstructured booklets where the same benefit is described differently by each carrier.

Grounded on

Booklet textStandard comparison template schema

Stays deterministic Template population and the email reply.

Human control

Analyst reviews the populated template before client discussion

Insurance Moderate risk

Policy Renewal Data Consolidation

Carriers submit daily renewal packages containing demographic data, experience summaries, rates, and claims across spreadsheet and PDF formats and multiple languages. Analysts extract the details and apply calculations to build a comparison report.

Pattern mapping

Document IntelligenceControlled Content Generation

GenAI opportunity

Normalises multi-language, multi-format carrier submissions onto one schema so they can be compared at all.

Grounded on

Carrier submissionsRate and claims schema

Stays deterministic Renewal calculations and database ingest.

Human control

Analyst reviews the consolidated report before renewal decisions

Insurance Low risk

Contract Classification and Routing

A small team manually reviews, categorises, and routes high-volume customer email across more than twenty categories. Transfer requests need additional portal lookups to identify the right administrator before forwarding.

Pattern mapping

Classification & RoutingAgentic Workflow

GenAI opportunity

Classifies across twenty-plus categories from varied customer wording, where maintaining rules for each category has proven unsustainable.

Grounded on

Category taxonomyHistorical routing decisions

Stays deterministic Subfolder moves, portal lookups, and forwarding.

Human control

Unclassifiable items flag as business exceptions for team review

Insurance Low risk

Claims Adjuster Assignment

Representatives download pending arbitration cases, search the claims system, identify claim type and exposure, verify whether payment has been made, assign an adjuster, and update the external forum.

Pattern mapping

Agentic WorkflowEntity Resolution & Matching

GenAI opportunity

Little. This process is deterministic end to end — lookup, rule, update. It is worth studying as a case where adding a model would add cost and failure modes without adding judgment.

Grounded on

Adjuster group definitionsClaim type codes

Stays deterministic All of it: portal navigation, assignment rules, reminders, and report generation.

Human control

Team reviews the summary report of assignments

Finance Critical risk

Revenue Contract Review

To recognise revenue, the finance team reads contract terms manually to establish which products were sold and the terms attaching to each line item. Contract and fee structures vary widely between agreements.

Pattern mapping

Document IntelligenceControlled Content Generation

GenAI opportunity

Interprets non-standard commercial terms line by line, where product and fee structures vary enough that no template covers them.

Grounded on

Contract textRevenue recognition policyProduct catalogue

Stays deterministic Population of the finance model and system-of-record lookups.

Human control

Finance analyst reviews extracted terms before revenue treatment is booked

Finance Critical risk

Financial Close

Closing the books is lengthy and manual, with organisations starting the next cycle as the current one finishes. The process is document-heavy and calculation-heavy, and matching commission lines to sales orders requires semantic judgment.

Pattern mapping

Data InvestigationAgentic WorkflowEntity Resolution & Matching

GenAI opportunity

Matches commission lines to sales orders semantically and explains the exceptions, which is the cognitive step blocking a repeatable close.

Grounded on

Commission statementsSales ordersOperating procedures

Stays deterministic Trial balance ingest, journal posting, and ledger lock.

Human control

Controller reviews preliminary close reports; exceptions handled by humans

Finance Low risk

Accounts Payable Enquiry Response

A shared mailbox receives several hundred vendor enquiries daily. Sorting is automated but response drafting remains entirely manual, producing long turnaround and strained vendor relationships.

Pattern mapping

Classification & RoutingInvestigation / Case ManagementControlled Content Generation

GenAI opportunity

Interprets what the vendor is actually asking and drafts the reply from invoice status, payment, and terms data.

Grounded on

Invoice databasePayment statusResponse templates

Stays deterministic Ticket creation and email metadata retrieval.

Human control

Requests lacking complete information raise a ticket for the AP team rather than auto-responding

Technology Moderate risk

IT Incident Investigation

During major incident calls, technical owners search historical incident and change tickets manually for relevant precedent. The search delays resolution and puts service level agreements at risk.

Pattern mapping

Knowledge & ResearchInvestigation / Case Management

GenAI opportunity

Summarises related historical incidents, recent changes, and runbook steps into a brief short enough to be useful during a live call.

Grounded on

Historical incident ticketsChange ticketsRunbooksKnowledge base

Stays deterministic Ticket retrieval and linking.

Human control

Technical owner remains the decision maker throughout

Technology Low risk

IT Service Request Triage

Requests arrive through voicemail, email, and chat and are categorised by hand before any resolution begins. Simple, repetitive request types consume service-desk capacity that would be better spent on outage prevention.

Pattern mapping

Classification & RoutingAgentic Workflow

GenAI opportunity

Interprets free-text and transcribed voicemail requests into catalogue request types.

Grounded on

Request catalogueHistorical ticket resolutions

Stays deterministic Assignment and access provisioning, which stay under existing entitlement controls.

Human control

Anything outside known request types routes to a human queue

Technology Moderate risk

Security Framework Gap Analysis

New regulatory and standards-body requirements are published periodically and need deep analysis to establish whether internal controls already satisfy them. The comparison is slow and easy to perform inconsistently.

Pattern mapping

Knowledge & ResearchDocument IntelligenceControlled Content Generation

GenAI opportunity

Compares new requirement text against internal control descriptions and articulates where coverage falls short — a semantic comparison with no structured equivalent.

Grounded on

Published regulatory requirementsInternal control documentation

Stays deterministic Report generation and distribution.

Human control

IT analyst reviews the gap analysis and owns remediation decisions

Technology Moderate risk

Vendor Onboarding

Onboarding stalls on incomplete documentation, repeated follow-ups, compliance and risk checks, and inconsistent data across systems. Approval bottlenecks and a poor vendor experience follow.

Pattern mapping

Document IntelligenceAgentic Workflow

GenAI opportunity

Reads varied vendor documentation and flags inconsistencies that would otherwise surface as a follow-up cycle weeks later.

Grounded on

Vendor-submitted documentsCompliance checklistScreening results

Stays deterministic ERP record creation and approval routing.

Human control

Legal and risk reviews remain human approval gates

Operations High risk

Pension Participant Address Reconciliation

A discrepant address arises when a former employee's address is wrong or mail is returned. These must be resolved quarterly under labour regulations, working through internal records, locator tools, and outreach, with every effort documented to evidence good faith.

Pattern mapping

Entity Resolution & MatchingData InvestigationAgentic Workflow

GenAI opportunity

Reconciles conflicting addresses across systems and locator results, and judges which is most likely current.

Grounded on

Internal HR and pension recordsThird-party locator resultsPrior outreach log

Stays deterministic Data normalisation, outreach dispatch, and the good-faith documentation trail.

Human control

Unmatched cases go to human review before outreach or lost-participant classification