> ## Documentation Index
> Fetch the complete documentation index at: https://developers.argosidentity.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Policy Configuration

> Define verification policies in natural language to create automated workflows.

## What is a Policy?

A policy is a set of **verification rules written in natural language** that defines what Omni should check and how it should evaluate documents. Omni's NLP Workflow Generator converts your policy into an executable verification pipeline.

## Writing a Policy

Policies are written as plain-text instructions that describe your verification requirements.

**Example policy for KYB:**

```text theme={null}
Verify the business entity by:
1. Extracting company information from registration documents
2. Screening the company name and directors against AML business watchlists
3. Validating business registration numbers against official registries
4. Cross-referencing beneficial ownership information
```

## Policy Best Practices

<AccordionGroup>
  <Accordion title="Be specific about verification steps">
    Clearly list each verification action. The more specific your policy, the more accurate the engine selection and orchestration.
  </Accordion>

  <Accordion title="Define pass/fail criteria">
    Specify what constitutes a successful verification versus a failure. This helps Omni's AI Decision Making engine produce clear Approve/Reject/Flag outcomes.
  </Accordion>

  <Accordion title="Reference document types explicitly">
    Mention the exact document types you expect (e.g., "passport", "business registration certificate") so Omni can optimize engine selection.
  </Accordion>
</AccordionGroup>

## How Policies Are Executed

<Steps>
  <Step title="NLP Parsing">
    Your natural language policy is parsed by the NLP Workflow Generator.
  </Step>

  <Step title="Engine Mapping">
    The system maps each verification step to the appropriate AI engines from the engine catalog.
  </Step>

  <Step title="DAG Construction">
    The Identity Agent constructs a Directed Acyclic Graph (DAG) that determines the optimal execution order.
  </Step>

  <Step title="Execution">
    When analysis is triggered, the engines execute in the determined order, passing results between steps.
  </Step>
</Steps>