Running Agentic Fraud Reasoning
This guide explains the requirements for running Agentic Fraud Reasoning, how to test an Orchestrator Agent, and what admins see while a document is being processed.
For initial setup, see Quick start.
Requirements for analysis to run
Before Agentic Fraud Reasoning can run on a document, the document must match an available orchestrator setup.
The main requirements are:
| Requirement | Description |
|---|---|
| Active Orchestrator Agent | An orchestrator must be available and active. |
| Assigned document type | The document type must be assigned to the orchestrator. |
| Assigned Fraud Analysis Agents | The orchestrator should have relevant Fraud Analysis Agents connected to it. |
| Submitted document | The document must be available for analysis. |
If these requirements are not met, Agentic Fraud Reasoning may not run for the document.
Automatic document analysis
When automatic fraud reasoning is enabled, Bynn can analyze submitted documents that meet the current orchestrator configuration.
Normal analysis does not need to be started manually from the document detail view. The applicable Orchestrator Agent is selected as part of the document analysis process.
Document type matching
Document type matching determines which Orchestrator Agent is used for a document.
For example:
- a passport document can use an orchestrator assigned to Passport
- a bank statement can use an orchestrator assigned to Bank Statement
If the document type is not assigned to an orchestrator, no fraud reasoning result may be created for that document.
To change which documents are analyzed, update the document types assigned to the relevant Orchestrator Agent.
Testing an orchestrator
Admins can test an Orchestrator Agent from the Orchestrator Agents page.
Testing an orchestrator is different from normal document analysis.
| Flow | Where the result appears |
|---|---|
| Test Orchestrator | Inside the Test Orchestrator modal. |
| Normal document analysis | In the document detail view, under Fraud Reasoning Analysis. |
Use the test function to check how an orchestrator performs against a document before relying on it for live document review.
Testing is useful when:
- a new orchestrator has been created
- document types have been changed
- Fraud Analysis Agents have been added or removed
- the orchestrator configuration has been updated
- an admin wants to test the orchestrator against a specific document
Using a Document ID for testing
When testing an orchestrator, the Test Orchestrator modal allows admins to enter a specific Document ID.
The Document ID field is optional.
- Enter a Document ID to test the orchestrator on a specific document.
- Leave the field blank to use the most recent document.
After clicking Run Test, the result is shown inside the modal.
The test result does not create a normal document detail result and does not appear as a standard fraud reasoning result on the document view.
Document processing state
While a document is being processed, its document card shows an AI Analysis label and a progress indicator.
This progress represents the broader document analysis process, including the different analysis steps performed by Bynn. It does not represent Agentic Fraud Reasoning alone.
The AI Analysis label is shown only while processing is in progress. After processing completes, the document card updates and the completed analysis views become available.
After normal analysis completes
After normal analysis completes, open the document detail view and select the Fraud Reasoning Analysis tab to review the result.
The document view also includes separate areas such as Risk Analysis and Applicant. These sections contain other document and applicant information and are separate from the Agentic Fraud Reasoning result.
For a detailed guide to the result interface, see Viewing fraud analysis results. For an explanation of each result field, see Understanding fraud analysis results.
If no Fraud Reasoning Analysis appears
If no fraud reasoning result appears for a document, Agentic Fraud Reasoning may not have run.
Common reasons include:
- no active orchestrator is available
- no orchestrator is assigned to the document type
- the document type is not supported by the current orchestrator setup
- the orchestrator has no relevant Fraud Analysis Agents assigned
- Agentic Fraud Reasoning is not enabled or available in the workspace
If a document type is not assigned to any orchestrator, normal analysis can be skipped without showing a document-level warning.
Admins may see a warning in the relevant settings area if some document types are not assigned to an orchestrator.
To resolve this, check the Orchestrator Agent configuration and confirm that the document type is assigned to an active orchestrator.
Running another orchestrator test
To test an orchestrator again:
- Open the Orchestrator Agents page.
- Find the orchestrator you want to test.
- Click Test.
- Enter a specific Document ID, or leave the field blank to use the most recent document.
- Click Run Test.
The new test result is shown inside the Test Orchestrator modal.
Running another test does not replace or update the normal Agentic Fraud Reasoning result shown in the document detail view.