Before you start
A signed-in NodeFox workspace. Use synthetic data and prepare outputs before enabling external writes.
01Choose a supported input
Use Reader File, Text or CSVExcel according to the input and your installed version. Test the actual file format with a non-sensitive sample. A successful file open does not guarantee useful text extraction from scanned or malformed content.
02Preserve the source
Keep a document identifier and clause, page or row reference with extracted values. For the renewal example, the relevant source is Clause 8.2, not a detached number returned by a model.
03Interpret then validate
Use Conversation for bounded extraction, then Code to validate required fields and apply exact checks. If extraction fails, produce a review item with the failure reason instead of a plausible-looking value.
04Prepare a useful artifact
Use Writer to prepare a report or structured export containing the source reference, extracted terms, checks and reviewer action. Keep acceptance, signing or posting separate from preparation.
If the result is wrong
Scanned PDFs and provider document support vary. Verify your actual reader/model combination instead of assuming universal OCR.