Anti-Money Laundering Agent Tool Invocation Matrix
Map function dependencies and routing logic for real-time transaction monitoring AI agents.
Use this template when architecting automated anti-money laundering (AML) investigation agents that query multiple compliance databases. It sets execution sequences, latency budgets, and fallback tools.
Role: Lead Financial Fraud AI Engineer specializing in deterministic tool routing and AML compliance.
Context
- Payment Infrastructure: {{payment_provider}}
- Investigation Alert Type: {{alert_category}}
- Latency SLA: {{latency_sla_ms}}
- Sanctions List Service: {{watchlist_service}}
- Internal Account Ledger API: {{account_ledger_api}}
- Alert Severity Threshold: {{alert_severity_level}}
Task
Produce an AML tool orchestration and dependency matrix defining how an autonomous investigation agent queries internal ledgers, sanctions lists, and external graph databases to assemble a SAR (Suspicious Activity Report) readiness packet.
Method
- Identify all analytical tools needed to resolve an alert for {{alert_category}}.
- Order tool calls into deterministic stages: triage, identity enrichment, graph expansion, and risk scoring.
- Map dependencies between tools to identify operations suitable for parallel execution versus sequential execution.
- Allocate fractional time budgets to each tool call to guarantee compliance with the overall {{latency_sla_ms}} limit.
- Define secondary fallback endpoints when {{watchlist_service}} or {{account_ledger_api}} encounters timeouts.
- Establish data transformation rules between the output of preceding tools and input schemas of subsequent tools.
- Structure the entire pipeline into an operational routing matrix.
Constraints
- MUST abort automated execution and route directly to human compliance analysts if {{alert_severity_level}} is exceeded.
- MUST NOT execute unindexed database queries that violate the {{latency_sla_ms}} budget.
- Primary sanctions lookups must query {{watchlist_service}} before ledger history is aggregated.
- All database modifications must be strictly disabled; agent tools must operate read-only.
Output format
1. Investigation Pipeline Topology
A concise summary of the end-to-end investigation workflow (150-200 words).
2. Tool Orchestration Matrix
A Markdown table containing the following columns: | Step Order | Function Name | Target System | Execution Type (Sync/Async) | Max Latency Budget | Fallback Strategy | (Provide 6 to 8 sequenced tool entries covering KYC lookups, balance checks, entity graph queries, and scoring).
3. Timeout and Error Protocol
A 3-step action protocol for handling endpoint outages during an active investigation.
Self-review
- Ensure the sum of individual latency budgets does not exceed {{latency_sla_ms}}.
- Check that read-only access constraints are clearly specified across all tools.
- Confirm that escalation logic triggers correctly when {{alert_severity_level}} is met.
Explicit role, a named task, and discrete steps the model can follow.
Background, inputs and variables the model needs before it starts.
Hard boundaries — what the model must and must not do.
A named, field-level shape for the response.
Ordered work items that force analysis before an answer.
Length and structure that travel across frontier models.
Signal density — instruction weight without padding.
Documented variables so the scaffold adapts to new inputs.
Quality bar, assumptions and behaviour when inputs are thin.
How much real usage the template has behind it.