Multi-Agent Handoff Drops Jurisdiction Flag Between Account-Opening and Compliance-Screening Agents

Issue: An Account-Opening Agent Notes in Free Text That a New Client’s Stated Residency and the Jurisdiction Implied by Their Funding Source Do Not Match, but the Structured Client Profile Handed Off to the Compliance-Screening Agent Has No Field for a Jurisdiction Conflict, So the Client Is Screened Only Under Their Stated Residency’s Rules

Frequency: Occasional

Symptoms

• A client account is approved and screened under a single jurisdiction’s regulatory rules, even though the account-opening agent’s own notes flagged that the client’s stated residency and their funding source’s jurisdiction do not match
• The structured client profile handed off to the compliance-screening agent contains the stated residency field used to select applicable rules, with no field capturing the funding-source jurisdiction mismatch the account-opening agent’s notes raised
• Compliance-screening agents operating purely from the structured profile show a materially lower rate of applying the additional jurisdiction’s rules to clients with a noted funding-source mismatch than screening agents given the account-opening agent’s full intake transcript
• The jurisdiction conflict surfaces only during a later regulatory examination or a cross-border transaction review, by which point the account has been open and active under a single jurisdiction’s rule set
• Clients whose funding source sits in the same jurisdiction as their stated residence never expose this, since the single stated-residency field the schema tracks already answers the only question the screening agent asks; the gap opens specifically when a second jurisdiction, like Country B in the funding source, has rules of its own that nothing in the profile prompts anyone to check

Root Cause The client-profile schema carries exactly one residency value because rule selection has historically been a one-jurisdiction lookup: read the field, apply that jurisdiction’s rules. A funding source domiciled elsewhere doesn’t fit that lookup’s shape, so when the account-opening agent notices Country B’s bank account doesn’t match the Country A residency on file, the observation has no field to land in and stays in intake notes the screening agent’s rule-selection step was never built to read. The screening step then does exactly what its schema asks of it – looks up Country A, applies Country A’s rules – with no signal that a second jurisdiction’s documentation requirement was ever in play.

Example

Account-opening agent processes a new client whose stated residency is Country A, but whose funding source is a bank account domiciled in Country B, and notes in free text: "Funding source jurisdiction (Country B) does not match stated residency (Country A) -- may require dual-jurisdiction screening"
Account-opening agent records the client's stated residency in the structured profile field used for rule selection, with no field for the funding-source jurisdiction mismatch
Structured profile handed off to the compliance-screening agent shows only Country A as the applicable jurisdiction
Compliance-screening agent screens the client only against Country A's rules and approves the account
Country B's stricter source-of-funds documentation requirement, which would have applied given the funding source, is never triggered, and the gap surfaces during a cross-border transaction review eighteen months later

Key Statistics

Contributing Factors

• Structured client-profile schema captures a single stated-residency field used for rule selection, with no field for a funding-source jurisdiction or other jurisdiction-conflict signal
• Account-opening agent’s jurisdiction-mismatch reasoning is recorded only in free-text intake notes, with no structured escalation path into the compliance-screening agent’s input
• No mandatory dual-jurisdiction screening trigger fires when the account-opening agent’s free-text notes contain jurisdiction-mismatch language, since the screening agent’s logic does not parse those notes

Mitigation Strategies

1. Add a Jurisdiction-Conflict Field to the Client-Profile Schema: Require the account-opening agent to record any funding-source or other jurisdiction-conflict finding in a dedicated structured field that triggers multi-jurisdiction screening, rather than leaving it only in free-text intake notes
2. Screening Agent Cross-Checks Intake Transcript for Jurisdiction-Mismatch Language: Require the compliance-screening agent to scan the account-opening agent’s free-text notes for jurisdiction-mismatch language before finalizing single-jurisdiction screening, not just the structured stated-residency field
3. Mandatory Dual-Jurisdiction Screening on Unresolved Conflict: Automatically trigger screening under both the stated-residency jurisdiction and any flagged funding-source jurisdiction whenever the account-opening stage’s output contains an unresolved jurisdiction-conflict finding
4. Track Conflict-Field-Absent Approval Rate: Continuously measure how often a client with an account-opening jurisdiction-conflict note is nonetheless screened under only their stated residency when the profile schema lacked a conflict field

Metrics

• Rate of account approvals where the account-opening transcript contains unresolved jurisdiction-mismatch language not reflected in the structured client profile
• Time between an account approval and a later-discovered jurisdiction-conflict gap for the same client
• Dual-jurisdiction screening rate for clients with a noted funding-source mismatch, segmented by presence vs. absence of a structured conflict field

Alerts

• A client account is approved while the account-opening transcript contains unresolved jurisdiction-mismatch language → P1
• A client already onboarded is found via later examination or transaction review to have had an unresolved jurisdiction conflict at approval time → P1
• Conflict-field-absent approval rate across a rolling window exceeds the defined threshold → P2

References

• Why Do Multi-Agent LLM Systems Fail? (MAST)
• Demystifying the Lifecycle of Failures in Platform-Orchestrated Agentic Workflows
• Agentic AI Systems Applied to tasks in Financial Services: Modeling and model risk management crews