AI Agents Reduce Financial Crime Investigation Backlogs for 900-Employee Firm

The Client Profile

• Industry: Financial Services

• Headquarters: Chicago, Illinois

• Company Size: Approximately 900 employees

• Engagement Focus: AI-powered fraud detection, KYC verification, and AML investigation automation

• AI Solution: Anthropic Claude

A Chicago-based financial services company serving commercial and consumer clients needed a more efficient way to manage growing fraud, Know Your Customer (KYC), and anti-money laundering (AML) workloads.

With approximately 900 employees and a rapidly expanding customer base, the organization processed thousands of account applications, identity documents, transactions, and compliance alerts each month. Its risk and compliance teams were responsible for investigating potentially suspicious activity while maintaining strict standards for accuracy, security, auditability, and regulatory oversight.

The Challenge

Growing Compliance Workloads and Manual Investigations

Fraud alerts and compliance cases were distributed across transaction-monitoring systems, customer databases, document repositories, spreadsheets, email, and third-party verification tools.

Analysts had to manually collect and review:

• Customer identification documents

• Account applications and onboarding records

• Transaction histories

• Beneficial ownership information

• Previous alerts and investigations

• Internal policies and risk criteria

• External watchlist and verification results

• Communications associated with the account

Each investigation required analysts to move between multiple systems, reconstruct timelines, compare records, document findings, and determine whether the case should be cleared or escalated.

This fragmented process created several operational challenges:

• Lengthy fraud and AML investigations

• Growing alert backlogs

• Inconsistent case documentation

• Duplicate research across compliance teams

• Delays during customer onboarding

• Limited visibility into investigator workloads

• Difficulty demonstrating how decisions were reached

• Excessive time spent on low-risk false positives

The organization did not want AI making final compliance decisions. It needed a governed solution that could perform repetitive investigative work while keeping qualified analysts in control.

The Solution

Claude-Powered Financial Crime Investigation Agents

Quandary Consulting Group designed an AI-powered financial crime operations platform using Claude as the intelligence layer.

The solution connected the company’s customer, transaction, document, fraud-monitoring, and case-management systems through a secure orchestration framework. Claude analyzed the information available for each case, identified relevant risk indicators, and generated a structured investigation summary for human review.

Rather than replacing fraud and compliance analysts, the platform gave them an AI investigation assistant capable of gathering evidence, interpreting complex documents, and preparing cases for faster resolution.

Automated Alert Intake and Case Creation

When the transaction-monitoring or identity-verification system generated an alert, the platform automatically created a case and assigned it based on risk level, investigation type, and analyst availability.

The AI agent collected the relevant information from connected systems, including:

• Customer and account records

• Identity-verification results

• Transaction activity

• Related accounts and counterparties

• Previous alerts

• Supporting documents

• Applicable policies and procedures

This eliminated the need for analysts to begin every investigation with manual data gathering.

AI-Powered KYC Document Review

Claude reviewed onboarding and identity documents to identify missing information, inconsistencies, and potential risk factors.

The system could compare information across applications, identification documents, business records, and customer profiles to flag issues such as:

• Conflicting names or addresses

• Missing beneficial ownership information

• Inconsistent dates of birth

• Expired identification

• Unexplained changes to account details

• Documents requiring additional verification

• Information that did not match the customer’s stated profile

Cases with complete, consistent documentation moved forward more quickly. Exceptions were routed to compliance personnel for further review.

Fraud and Transaction Analysis

For fraud and AML alerts, Claude organized transaction activity into a clear chronological narrative.

The AI agent highlighted unusual behavior, including:

• Sudden changes in transaction volume

• Activity inconsistent with the customer’s profile

• Rapid movement of funds between accounts

• Unusual geographic patterns

• Repeated transactions below review thresholds

• Connections to previously investigated entities

• Transactions lacking an apparent business purpose

The system presented these indicators as evidence for analysts to evaluate—not as final determinations.

Automated Investigation Summaries

After reviewing the available information, Claude generated a standardized case summary that included:

• Reason for the alert

• Customer and account background

• Relevant transaction activity

• Identified risk indicators

• Supporting and contradictory evidence

• Missing information

• Applicable internal policies

• Recommended next steps

• Source references for analyst verification

Every conclusion was linked to the underlying record or document, allowing analysts to validate the information before taking action.

Human-in-the-Loop Decision Controls

Quandary incorporated mandatory human review into all consequential decisions.

Authorized analysts remained responsible for:

• Clearing fraud and AML alerts

• Restricting or closing accounts

• Requesting additional customer information

• Escalating high-risk investigations

• Preparing regulatory filings

• Approving final case dispositions

Low-confidence findings and conflicting information were automatically routed for deeper investigation. The system could recommend an action, but it could not independently make a regulatory or customer-impacting decision.

Governance, Security, and Auditability

The platform was designed around the organization’s security and compliance requirements; key controls included:

• Role-based access permissions

• Encryption for data in transit and at rest

• Approved data-source restrictions

• Human approval checkpoints

• Complete AI activity logs

• Source-level citations

• Version-controlled prompts and policies

• Confidence thresholds

• Automated exception routing

• Retention controls aligned with company policies

The organization could review what information Claude accessed, what it generated, which recommendations were accepted or rejected, and who made the final decision.\

Key Platform Capabilities

• AI-Powered Investigations: Claude gathered, analyzed, and summarized information across fraud, KYC, and AML systems.

• Intelligent Document Review: The platform extracted information and identified inconsistencies across onboarding and compliance documents.

• Automated Case Narratives: Analysts received structured summaries with transaction timelines, risk indicators, and supporting evidence.

• Policy-Grounded Analysis: Claude evaluated cases against approved internal procedures and risk criteria.

• Human Approval Workflows: Qualified employees retained control over every consequential compliance decision.

• Centralized Case Management: Alerts, documents, findings, approvals, and final dispositions were managed through one controlled workflow.

• Complete Audit Trails: Every automated and human action was recorded for internal review and regulatory examinations.

The Results

_Faster Investigations with Stronger Oversigh_t

The Claude-powered platform transformed how the organization managed financial crime compliance.

Fraud and compliance analysts no longer spent most of their time gathering records, switching between systems, and manually reconstructing transaction histories. Instead, they received investigation-ready cases with the most relevant evidence already organized for review.

The organization achieved:

• Faster fraud, KYC, and AML investigations

• Reduced alert backlogs

• More consistent case documentation

• Quicker onboarding for legitimate customers

• Stronger detection of cross-system risk indicators

• Greater analyst capacity without proportional headcount growth

• Improved visibility into case status and team performance

• More defensible compliance decisions

• Complete traceability across AI and human actions

By combining Claude’s analytical capabilities with governed workflows and human oversight, the company created a scalable financial crime compliance operation that improved productivity without compromising accountability.