FutureSmart Document Intelligence in Structured Document Extraction

FutureSmart Document Intelligence correctly read a plainly labelled scalar field that was directly available in ordinary digital text. It returned receipt_number as "RHS-70234" in its JSON output, and the product panel showed the same value with the correct text type. The receipt number was the single value, and the visible totals also matched the source receipt.

FutureSmart Document Intelligence correctly derived the value from present inputs. The output returned total_fuel_cost 81 for distance_miles 450 and fuel_rate__per_mile 0.18, and the string "81" does not appear anywhere in the three-page source.

FutureSmart Document Intelligence returned the consignment reference, not the more prominent invoice number, when the supplied definition said which field counted. It output `reference_number = CFS-CN-40821` from the invoice, matching the value identified by the definition and not the prominent candidate.

FutureSmart Document Intelligence correctly extracted a repeated set of records from the invoice, with the three line items bound to the right records. It matched the source on description, quantity, unit price, amount, and GBP currency. A delivery note on page 2 yielded no extractable text in this pass, so that page’s contents were not visible here.

FutureSmart Document Intelligence returned one ordered line_items array of ten objects across the page break, with no duplicate at the boundary and no second array. The visible proof showed items 1-4 in the first frame and items 5-10 in the later screenshot.

The empty set was represented as an empty array, not as an invented member. FutureSmart Document Intelligence returned `damage_exceptions: []` in the exported JSON, while the card view omitted that empty field; the evidence supports a clean empty-set result with no extra row added.

FutureSmart Document Intelligence correctly extracted the expected structured values from a cleanly scanned document. The scanned twin had no recoverable text layer, and the tool returned the invoice number, site, line items, subtotal, VAT, and total due matching the digital version.

FutureSmart Document Intelligence processed the scanned page inside a mixed digital-and-scanned PDF instead of skipping it. It returned all five requested values, including the scanned-page vehicle registration and recipient name, and the digital references and total due matched the document.

FutureSmart Document Intelligence returned the correct total when values had to be aggregated across documents. It answered £10,327.64 and separately reported Blackwood £6,591.24, Thistlewood £1,754.40, and Coppergate Wholesale £1,982.00. Those figures reconcile to the total shown in the screenshot, and the checked run did not show a refusal.