// Chapter 06
Review results & fix mistakes
AI extraction is good, not infallible — so typed extracted fields are editable in the Details tab. Corrections are validated and remembered.
Edit a field
Open the document's Details tab.
Typed extracted fields are listed there. A field the AI couldn't find shows a dash (—) — it's never hidden, because a blank you can fill is more useful than a blank you can't see.Click the pencil.
It appears when you hover (or tab to) a field. The field becomes an input matched to its type — a date picker for dates, a numeric input for amounts, Yes/No for checkboxes.Save with ✓ or Enter.
Cancel with ✕ or Escape. Each save records one correction. If the old value remains on screen, close and reopen the document to refresh the Details tab.
Inline correctionIllustrative demo
What happens behind the scenes
- Your value is validated, server-side, against the document type's schema. A value that doesn't fit (text in an amount field, say) is rejected with an inline explanation — bad data never lands.
- The document does not reprocess. Fixing a field never sends the document back through OCR or classification — your correction simply becomes the value, and the document stays
ready. - Every correction is remembered: who changed what, when, and from what to what — per document, for as long as the document exists.
Free-form document types (correspondence, reports, generic forms) show their extracted data as-is rather than as typed fields — the structured editing above applies to typed documents like invoices, contracts, and purchase orders.
That's the core loop: upload or email in, let matching rules file documents, and fix the rare miss. The remaining chapters cover the surrounding machinery — approvals, search, your team, and billing.