Financial Crime Network Analysis with AI — Graph Analytics, Shell Companies and Correspondent Banking

Network Analysis as the Core of Advanced Financial Crime Detection

The most sophisticated financial crime — sanctions evasion, kleptocracy, large-scale money laundering — is not caught by looking at individual transactions. A single wire transfer from Company A to Company B may appear entirely legitimate. The laundering is revealed only when you see the entire network: Company A is owned by Company B through a Cayman holding company, which is itself beneficially controlled by a sanctioned individual through a chain of nominees. The transaction is a round-trip — value is leaving and returning through a different legal entity.

Network analysis — graph-based mapping of relationships between entities, accounts, transactions, and individuals — is the analytical framework that reveals these structures. Historically, this was a resource-intensive manual process: investigators would spend weeks building relationship diagrams for a single complex case. AI-assisted network analysis compresses this dramatically. Claude can help investigators build entity maps, identify typology patterns, trace beneficial ownership chains, and generate investigation narratives from network data.

Beneficial Ownership Mapping

The FinCEN Customer Due Diligence Rule (CDD Rule, 31 CFR 1010.230) requires banks to identify and verify the beneficial owners of legal entity customers — the natural persons who ultimately own or control the entity at the 25% threshold. For complex structures (multi-layer LLCs, offshore holding companies, trusts, nominee arrangements), tracing ownership to the UBO is analytically demanding. Claude excels at mapping these structures from the documents that CDD collects.

• "Trace the beneficial ownership for the following entity structure to the 25% UBO threshold required by FinCEN's CDD Rule. Entity: Meridian Group Holdings LLC (Delaware). Managing member: Sunrise Capital Partners LLC (Cayman Islands, 100% owner). Sunrise is owned: 55% by John Davies (UK national, DOB 1971), 30% by Clearview International Foundation (a Liechtenstein foundation, beneficiaries not named), 15% by Pacific Holdings Ltd (Hong Kong, no individual UBOs provided). Which natural persons reach the 25% ownership threshold? What is the CDD/EDD status for each identified UBO? What additional documentation is needed for the Clearview Foundation and Pacific Holdings to complete the UBO identification? What is the overall risk tier for this structure?"
• "I have 8 LLCs in our bank's portfolio that appear to be related — they share the same registered agent, were incorporated within 3 weeks of each other in Delaware and Wyoming, and have had wire transfers between them totaling $4.2M in 6 months. None of them have provided complete beneficial ownership information. Map the potential network connections between these entities. What ownership structure investigation should I initiate? What additional documents should I collect? What are the SAR indicators if the UBO investigation reveals these are controlled by the same individual(s)?"

Shell Company and Layering Pattern Detection

Shell companies — entities with no real business purpose, employees, or economic activity, used purely to hold assets or move money — are the primary vehicle for layering in money laundering. FATF's typologies and FinCEN's advisories identify specific characteristics that distinguish legitimate holding companies from shell company vehicles for laundering. AI can help analysts systematically apply these typologies to their customer portfolio.

• "Apply FATF's shell company red flag typologies to this customer: Customer: Global Trade Services LLC (Wyoming, 2 years old). Registered agent: Registered Agents Inc (known high-volume registered agent). No employees, no physical office address (registered address is the agent's address). Business: 'international commodity trading.' Annual transaction volume: $28M in wire transfers. Wire recipients: 14 different companies in 8 different countries, none publicly known as commodity traders. Owner: a New Zealand national with no prior relationship to commodity trading in LinkedIn/public profiles. No website, no verifiable business presence. Apply the FATF red flags: (1) entity formation red flags, (2) transaction pattern red flags, (3) counterparty red flags, (4) owner/controller red flags. Calculate a composite shell company risk score."
• "Help me build a shell company screening model for our commercial banking portfolio. I want to score all 340 business accounts against a shell company risk factor matrix. Factors to include: (1) Registered agent type (high-volume vs. local law firm), (2) State of incorporation (Delaware, Wyoming, Nevada score higher), (3) Age of entity vs. transaction volume ratio, (4) Number of employees vs. transaction volume, (5) Business description specificity (generic vs. specific), (6) Website presence and quality, (7) Geographic concentration of counterparties, (8) Round-number transaction frequency, (9) Owner nationality vs. business location mismatch. Assign point values to each factor and define risk tier thresholds."

Correspondent Banking and High-Risk Corridor Analysis

Correspondent banking — where a domestic bank provides services to a foreign bank to allow the foreign bank's customers to access the domestic financial system — is a major vector for international money laundering and sanctions evasion. The correspondent bank has limited visibility into the transactions flowing through the nested account. FinCEN's guidance on nested accounts (FIN-2006-G010) and FATF's correspondent banking guidance require enhanced due diligence on correspondent relationships with banks in high-risk jurisdictions.

• "Analyze our correspondent banking portfolio for high-risk corridors. We have 23 correspondent relationships. Flag for enhanced review: any correspondent bank in a FATF grey-listed or blacklisted jurisdiction, any correspondent bank whose home country is subject to OFAC comprehensive sanctions programs, any correspondent bank with a history of regulatory sanctions (AML failures, OFAC violations) in the past 5 years. For flagged relationships: (1) characterize the risk (sanctions exposure vs. AML risk vs. both), (2) identify what EDD we should be performing per FinCEN's nested account guidance, (3) recommend a risk-tiered monitoring frequency (monthly / quarterly / annual review). Here are our 23 correspondents: [list]."
• "I need to build a transaction monitoring rule for nested account activity through our correspondent banking relationships. A 'nested account' is where a third bank (not our direct correspondent) uses our correspondent's account at our bank to access the US financial system. This creates opacity — we see the correspondent's account activity but not who is behind each transaction. Design a rule set to detect nested account patterns: (1) transaction volume inconsistent with the correspondent's stated business and customer base, (2) geographic origination of transactions inconsistent with the correspondent's home country, (3) unusual transaction timing or frequency patterns, (4) beneficiary concentration in high-risk countries. For each rule: specify the threshold, the alert logic, and the investigation step."

Link Analysis for Investigation Support

Link analysis — identifying and visualizing connections between entities in an investigation — is a standard tool in financial crime investigation. Intelligence agencies and law enforcement have used it for decades. Claude can help investigators build link analysis maps from case data and identify the most significant connections for further investigation.

• "Build a link analysis map for the following financial crime investigation. Case: suspected human trafficking proceeds laundering. Known entities: (1) Pacific Rim Entertainment LLC (nightclub, large cash deposits), (2) Global Wellness Holdings (spa chain, accounts at our bank), (3) Chen International Trading (wire transfers to Vietnam and Philippines), (4) Michael Chen (sole owner of all three entities). Known connections: Pacific Rim → Global Wellness: $340K monthly cash transfers, described as 'management fees.' Global Wellness → Chen International: $280K monthly wire transfers, described as 'supply chain payments.' Chen International → accounts in Vietnam and Philippines: $260K monthly, purpose unknown. Identify: (1) the likely placement-layering-integration structure, (2) the red flags specific to the human trafficking typology (FATF Report on Financial Flows from Human Trafficking), (3) additional accounts or entities I should subpoena to complete the network map."

Where to Start

Financial crime network analysis starts with beneficial ownership — the data you already collect under the CDD Rule. Use Claude to help map complex ownership structures more efficiently and consistently. The shell company risk scoring model is the next step: run it against your existing commercial portfolio quarterly to identify accounts that have aged into a higher risk profile. For specific investigations where you suspect layering through multiple entities, the link analysis prompts above can help organize and interrogate the case data. The AML & Financial Crime guide covers the SAR writing and regulatory examination workflows that follow a completed network investigation.