Landing AI in Structured Document Extraction

Landing AI passed this scenario by returning an honest null for the absent field. The metadata showed no source range for that field. The receipt number was returned as RHS-70234, with a source range shown in the metadata.

Landing AI passed this scenario. It returned the correct owed amount from three documents with different layouts: $90.10 from the receipt, $227.33 from the invoice, and USD 157.50 from the other receipt. The screenshots show each value highlighted in the viewer and the same single String field in the schema.

Landing AI derived the requested value in this scenario. The invoice shows DISTANCE 450 miles and FUEL RATE £0.18 per mile; the JSON returns fuel_cost as GBP 81.00, and the document text does not contain 81.00. The playground screen shows the same result beside a one-field schema.

Landing AI extracted the repeated set of records in this run. The invoice table on page 3 returned three line items with the correct values and pairing; the delivery note says there were three lines and six units, and the schema asked for description and amount only.

Landing AI left the non-record out of the extracted set in the tested invoice. The output contained five members, matching the five numbered rows, and it did not include the separate Total row or its 1,982.00 amount. The schema description shown in the capture was neutral.

Landing AI kept the repeated set in one array when the data crossed a page break. In the checked case, ten line items appeared in document order with no duplicate at the boundary, and the results panel showed the array opening once and closing once. No other case is reported here.

Landing AI returned an empty set when the document said there were none this period. In the checked PDF page, the source text said no damage or exceptions were reported, and the export showed an empty array. The schema pane showed the field as a required array, and the output did not turn the marker text into a member.

Landing AI produced accurate text for the scanned-document case tested here. The digital invoice and its image-only twin returned the same three values: invoice_number TPH-3301, invoice_date 10 March 2026, and total_due £1,754.40. The twin file is shown with zero extractable characters, and the gap was none. No other test case has a result yet.

Landing AI failed this mixed PDF scenario: it returned the scanned-page values and the digital line items correctly, but total_due came back as “#1,246.80” instead of “£1,246.80”.

Landing AI failed to total values across every document in the project. The shown schema returned each file's own total instead of one cross-document total, and chat replied that it could not find the answer in the provided document. Blackwood showed £6,591.24 and HIS-4402 showed £1,246.80; Thistlewood is shown as outdated under the current schema.

Landing AI did not filter records across the document set in this test. The chat reply said it could not find the answer in the provided document, and the results were shown one file at a time. The extracted values still matched their source invoices.

Landing AI answered a document question outside the extraction schema and tied the answer to the source document. In the invoice test, it returned 06 Jul 2020 for the due date and showed a visual reference to the row containing that value. A second wording returned the same attributed answer.

Landing AI passed this scenario by returning an honest null for the absent field. The metadata showed no source range for that field. The receipt number was returned as RHS-70234, with a source range shown in the metadata.