Workflow boundary
DocsRouter provides OCR. It does not bundle a claim, assess damage, determine fault, detect fraud, map ACORD data, or authorize payment.
Use /v1/ocr to extract one supported image or document at a time. For multi-page reports and estimates, inspect result.pages and validate that the expected pages and fields are present.
Your claims application can then normalize the OCR result into its internal schema, compare it with policy and incident data, and route exceptions to an adjuster.
Keep photo-damage analysis, liveness checks, fraud models, claim decisions, and core-platform integration as separate, tested services. Do not infer those capabilities from a vision model's ability to read text.