Flagging entries whose written justification is inadequate before they reach an approver or an auditor.
Language model evaluating narrative adequacy against policy criteria
LLM API