Automating wire and payment remittance AML triage with the Splunk platform and Jev
Financial institutions process high volumes of wire and electronic payments daily across networks like SWIFT, Fedwire, and ACH. Compliance teams must screen these transactions against anti-money laundering (AML) regulations and global sanctions lists.
A persistent challenge lies in unstructured free-text fields, such as remittance_info or payment reference notes. Traditional keyword-matching rules, such as the lookup-based screening in Wire transfers into suspicious or banned countries, frequently generate high volumes of false positives (such as flagging "Cuba Street" in New Zealand as a Cuban sanctions violation) while missing nuanced evasions or references to dual-use goods.
This use case helps organizations that want to automate the contextual evaluation of unstructured remittance text alongside structured payment attributes. By scoring transactions with natural language inference (NLI), compliance teams can automatically clear benign payments and prioritize high-risk transfers for immediate Tier 2 investigation.
How Jev evaluates payments
Jev is an AI model from TypeSafe, designed for natural language inference and contextual evaluation. The Jev for Splunk app adds a jev custom search command that sends field values from your search results to the Jev API and writes the returned judgments back as fields, so you can run the evaluation from within your search without any custom scripts.
Traditional detection methods rely on static regex or keyword lookups, which lack semantic understanding. The jev command passes structured event fields and unstructured free text to the Jev model, which assesses whether the combined context satisfies a specific natural language condition. It evaluates the relationship across entities, transaction sizes, and remittance memos together.
Payment memos often defeat standard keyword filters in two ways:
- False positives (keyword traps): A wire containing the memo "Lunch catering for Cuba St office meeting Wellington" triggers traditional sanctions filters because of the word "Cuba", even though the context clearly indicates a local restaurant on Cuba Street in New Zealand. Jev evaluates the full sentence context alongside the sender and beneficiary entities to recognize benign activity.
- False negatives (nuanced evasions and dual-use goods): A memo stating "Invoice 8841-A dual-use centrifugal pump parts batch 3" or "Daily cash drop structured transfer sequence 4 of 6" might not hit simple sanctions list names, but Jev can recognize the regulatory significance of dual-use export controls and of structured transactions designed to evade reporting thresholds. For rule-based searches that detect structuring, see Monitoring money laundering activities with the Splunk platform.
The jev command evaluates event data by transmitting the specified fields over secure HTTPS to the TypeSafe API (api.typesafe.ai). While TypeSafe's policy specifies that customer queries and responses are not used to train models, payment remittance and entity fields often contain personally identifiable information (PII) and sensitive financial details. Before you deploy this solution:
- Review data governance policies: Make sure your InfoSec, compliance, and data privacy teams approve the transmission of wire and payment memo attributes to third-party inference services.
- Verify data residency: Confirm that the external API endpoints align with your organization's jurisdictional and cross-border data transfer requirements.
- Practice data minimization: Restrict the
stateargument to only the fields necessary for context evaluation.
Data required
Financial data for wire transfers, with these fields:
transaction_id: Unique identifier for the payment.amount: Numerical transfer amount.currency: Currency code, for example USD or EUR.sender_name: Name of the ordering customer or originating institution.beneficiary_name: Name of the ultimate recipient or entity.remittance_info: Free-text field describing the purpose of payment, memo, or invoice details.
Prerequisites
- Splunk Enterprise or Splunk Cloud Platform, version 9.2 through 10.5
- The Jev for Splunk app, installed from Splunkbase. This is an unofficial, developer-supported app. It isn't an official TypeSafe product.
- Access to the Jev API, with a TypeSafe API key
- A role that holds the
run_jevcapability
How to use Splunk software for this use case
Configure the Jev for Splunk app
Open the Jev for Splunk app, and on the Configuration page select the Account tab. In the TypeSafe API key field, enter your API key, then click Save. All other settings use their default values.
The Account tab includes the TypeSafe API key field, along with the Base URL and Model settings that you can leave at their defaults.

Inspect your wire transfer data
In this example, wire transfer data from a banking institution's payments is stored in the payments index. Run the following search to look at the data:
index=payments sourcetype=wire:transfer
Each event includes the payment fields that the scoring search evaluates.

Score and triage wire transfers
Run the scoring search on your live wire event indexes, over a time window that suits scheduled triage alerts. The search returns a probability score for each of two questions, one about sanctions and one about AML, and then categorizes each transfer into an actionable triage tier (triage_disposition) for routing.
index=payments sourcetype=wire:transfer earliest=-15m@m latest=@m
| eval sanitized_remittance = replace(remittance_info, "(?i)\b\d{8,17}\b", "[ACCOUNT_REDACTED]")
| jev state="transaction_id,amount,currency,sender_name,beneficiary_name,sanitized_remittance" noul="Does this payment involve a sanctioned entity, restricted jurisdiction, or dual-use good?" as=sanctions_probability
| jev state="transaction_id,amount,currency,sender_name,beneficiary_name,sanitized_remittance" noul="Does this payment show indicators of money laundering, structuring, or deceptive remittance purpose?" as=aml_probability
| eval triage_disposition = case(sanctions_probability >= 0.50, "Escalate_Sanctions_Block", aml_probability >= 0.65, "Escalate_AML_Investigation", sanctions_probability < 0.20 AND aml_probability < 0.20, "Auto_Clear",true(), "Standard_Review")
| fieldformat amount = "$".tostring(amount,"commas")
| table transaction_id, amount, remittance_info, sanctions_probability, aml_probability, triage_disposition
The following table explains what each part of the search does:
| Splunk Search | Explanation |
|---|---|
index=payments sourcetype=wire:transfer earliest=-15m@m latest=@m |
Search your wire transfer data over the last 15 minutes, snapped to the minute, to match a 15-minute alert schedule. |
| eval sanitized_remittance = replace(remittance_info, "(?i)\b\d{8,17}\b", "[ACCOUNT_REDACTED]") |
Create a sanitized_remittance field that replaces any 8 to 17 digit number in the remittance text with [ACCOUNT_REDACTED], so account numbers aren't sent to the API. |
| jev state="..." noul="Does this payment involve a sanctioned entity, restricted jurisdiction, or dual-use good?" as=sanctions_probability |
Send the listed fields to Jev with a yes/no sanctions question, and store the probability that the answer is yes in the sanctions_probability field. |
| jev state="..." noul="Does this payment show indicators of money laundering, structuring, or deceptive remittance purpose?" as=aml_probability |
Send the same fields to Jev with a separate yes/no AML question, and store the result in the aml_probability field. |
| eval triage_disposition = case(...) |
Assign each transfer a triage disposition based on the two scores, using the thresholds in the table in Interpret the results. |
| fieldformat amount = "$".tostring(amount,"commas") |
Display the amount with a currency symbol and commas, without changing the underlying value, so the column still sorts by number. |
| table transaction_id, amount, remittance_info, sanctions_probability, aml_probability, triage_disposition |
Display the results in a table. |
The search results show the sanctions_probability and aml_probability scores and the triage_disposition assigned to each transfer.

Interpret the results
When you review the results, focus on three fields:
sanctions_probability: The probability that the answer to the sanctions question is yes.aml_probability: The probability that the answer to the AML question is yes.triage_disposition: The disposition of the wire transfer, based on thesanctions_probabilityoraml_probabilityscore.
Jev returns a decimal value between 0.00 and 1.00 for both sanctions_probability and aml_probability. Based on these scores, the search classifies each transaction into an actionable triage tier:
| Triage disposition (triage_disposition) | Trigger condition or threshold | Risk level | Recommended triage action | Practical example |
|---|---|---|---|---|
Escalate_Sanctions_Block |
sanctions_probability >= 0.50 |
Critical / Sanctions | Route to the OFAC or sanctions queue for an immediate payment hold. Trigger a downstream alert action or Splunk SOAR workflow. | Explicit references to dual-use industrial parts or restricted jurisdictions, for example "Invoice 8841-A dual-use centrifugal pump parts batch 3" |
Escalate_AML_Investigation |
aml_probability >= 0.65 |
High / AML risk | Assign to the Tier 2 AML queue for SAR/STR analysis. Trigger a downstream ticketing or Splunk SOAR workflow. | Stated structuring or deceptive payment flows, for example "Daily cash drop structured transfer sequence 4 of 6" |
Auto_Clear |
sanctions_probability < 0.20 AND aml_probability < 0.20 |
Low / Negligible | Eligible to bypass manual queues or fast-track straight-through processing. | Routine operational expenses and keyword traps, for example "Lunch catering for Cuba St office meeting Wellington" |
Standard_Review |
All other score combinations | Moderate / Ambiguous | Queue for Tier 1 manual review and standard customer due diligence (CDD). | Large commercial transfers, escrow deposits, or high-value crypto OTC desk settlements without overt red flags |
Teams can route on the triage_disposition field:
Escalate_Sanctions_Block: Flags high-risk sanctions matches for immediate payment holds and priority routing to OFAC or sanctions queues, or to automated Splunk SOAR playbooks.Escalate_AML_Investigation: Flags suspicious structuring or AML indicators for assignment to Tier 2 investigators or automated SAR/STR review workflows.Standard_Review: Flags moderate-risk or ambiguous transfers for Tier 1 manual review and routine customer due diligence.Auto_Clear: Identifies benign transfers eligible for automated clearing or fast-track processing, to reduce manual analyst queues.
The thresholds shown in the table above are illustrative starting baselines, defined by your institution's risk policy rather than by the model. Risk tolerance, payment volumes, and memo formats vary across institutions, compliance programs, and models. Governance teams should backtest and calibrate these cut-offs against historical payment logs and known false positives before turning on automated routing or Auto_Clear actions in production.
With these three fields, you have a classifier that can help your team triage transfers and decide which ones require further review, and at what level. As data volumes grow, more classification questions like these arise, and with the Splunk platform and Jev you can answer them where the data already lives.
Schedule the triage search as an alert
- Save the search as a scheduled alert that runs on the cron schedule
*/15 * * * *(every 15 minutes), so it evaluates data over the matching rolling window (earliest=-15m@m latest=@m). - Set the trigger condition to fire when the number of results where
triage_disposition != "Auto_Clear"is greater than 0. - Configure downstream alert actions, such as routing high-risk transfers (
Escalate_Sanctions_BlockorEscalate_AML_Investigation) to Splunk SOAR playbooks, webhook endpoints, or compliance investigation queues.
For step-by-step guidance on setting up scheduled alerts, see Create scheduled alerts.
Next steps
To learn more about how to detect financial crime and automate AML triage, contact your Splunk sales representative today. You can contact your account team through the Contact Us page.
You might also find the following resources useful for implementing the guidance in this article:
- Splunk Lantern Article: Querying payment rail logs in natural language using Splunk DSDL's LLM-powered chat UI
- Splunk Lantern Article: Leveraging generative AI capability in security operations with the AITK
- Splunk Lantern Article: Creating, monitoring, and optimizing LLM retrieval augmented generation patterns
- Splunk Lantern Article: Using the Cisco Time Series Model 1.0 on DSDL 5.2.3
- Splunk Lantern Article: Using generative AI to write and explain SPL searches
- Splunk Lantern Article: Using modern methods of detecting financial crime
- Splunk Lantern Article: Detecting wire transfer fraud
- Splunk Blog: Faster insights with third-party LLM services in Splunk search
- Splunk Blog: Talk to your logs: LLM-powered chat UI in DSDL 5.2.3
- Splunk .conf: Integrating GenAI with Splunk to drive digital transformation
- TypeSafe Docs: API reference

