Overview
Document AI produces a status for each uploaded document from two layers of rules, and the per-document statuses combine into the feature status using the precedence Declined > In Review > Approved.- Basic rules — built-in checks, each mapped to a configurable action:
DECLINE,REVIEW, orNO_ACTION. - Custom status rules — your own conditions on any extracted field, including cross-references against other steps.
Basic rules
Each rule below is detected automatically during extraction and forensic analysis. Configure the action for each in the Document AI feature’s Rules tab.| Rule | Config field | Triggers when |
|---|---|---|
| Document unreadable | document_ai_unreadable_document_action | The document can’t be read or no data can be extracted |
| Missing required fields | document_ai_missing_required_fields_action | A field marked required can’t be extracted |
| Document tampering | document_ai_document_tampering_action | Forensics detect signs of manipulation |
| Name mismatch | document_ai_name_mismatch_action | The holder name on the document doesn’t match the verified identity (below document_ai_name_match_score_threshold) |
| Unsupported file | document_ai_unsupported_file_action | The uploaded file type is not supported |
| Exceeded attempts | document_ai_max_attempts_exceeded_action | The user exceeds document_ai_max_retry_attempts |
document_ai_name_match_score_threshold.
Custom status rules
Custom status rules let you set a status (Approved, In Review, Declined, or No Action) when a condition on an extracted field is met. Each rule is:
value_type: "field" and point value at another field path:
Declined/In Review back up. No Action rules are recorded but leave the status unchanged.
Extracted fields are addressable as
document_ai.<field_key> using the key you configured for each field. Number and date fields support comparison operators (greater_than, less_than, date ranges); text fields support equality, contains, and fuzzy_match.