Named fields
Ask for the values your next step needs, rather than assuming a full page is the right delivery format.
AI DOCUMENT DATA EXTRACTION
Extract the PDF fields your Excel workflow already uses.
Turn existing column names into plain-language field requests, keep the same order in the exported workbook, and compare every value with its document source.
A useful output begins with its shape.
A document page is not automatically a record. State what should become a field, what belongs to one record, and what a person must review.
Explicit records
Define whether a line, section, or document becomes one output record before checking its values.
Source review
Keep the document beside the output when a field will affect a business, financial, or compliance workflow.
PUBLIC FIELD EVIDENCE
Four field definitions tied to one inspectable source.
The rows shown in the interactive demo are declared source values. They illustrate destination-column shape only and must not be read as AI output, customer data, or a general accuracy result.
Public synthetic invoice
Three synthetic lines, six printed columns, no personal data, and a declared reference workbook for side-by-side review.
Line description
Source column: DescriptionKeep printed wording and the invoice-line boundary together.
HS code
Source column: HS codePreserve every digit that is visible in the source instead of inferring a replacement.
Quantity
Source column: QuantityCheck the unit and decimal format before a downstream calculation uses it.
Line total USD
Source column: Line total USDCheck the currency and printed amount against the source document.
EXISTING EXCEL COLUMN WORKFLOW
Turn the columns your team already uses into plain-language field requests.
FOVATA does not import an Excel template on this screen. Use the header names and business rules from your current workbook to define fields in the same order, then export a new workbook and review it before copying or importing data downstream.
Read your header row
Choose only the columns the next process actually needs.
Describe each field
State the source meaning, missing-value rule, and format in ordinary language.
Keep the same order
The exported single-value sheet follows the configured field order.
Review before import
Compare every row with its source before it enters another system.
PLAIN-LANGUAGE SETUP
Eight workbook columns, eight reviewable instructions
No JSON download or hidden template file is required. These are the exact kinds of natural-language instructions a user can enter and save in the signed-in workflow.
- 01
Source page
Record the PDF page that supports this row so a reviewer can return to the source.
- 02
Vendor
Use the supplier name printed as the invoice issuer; do not substitute the buyer or ship-to party.
- 03
Invoice no.
Copy the invoice identifier exactly, preserving letters, digits, spaces, and punctuation.
- 04
PO no.
Return the purchase-order reference when printed; leave it blank when the document does not provide one.
- 05
Invoice date
Use the printed invoice date and normalize it as YYYY-MM-DD only when the date is unambiguous.
- 06
Due date
Use the explicit payment due date, not a date inferred from payment terms.
- 07
Currency
Return the printed ISO currency code or symbol context; flag an ambiguous currency for review.
- 08
Total
Return the final invoice total in the stated currency, excluding subtotal and tax-only amounts.
DECLARED REFERENCE OUTPUT
Different pages, one stable Excel row shape
Select a synthetic invoice page to see how its declared reference values fit the same destination columns. This is not live or historical model output.
| Source page | Page 1 |
|---|---|
| Vendor | Northwind Test Supply |
| Invoice no. | NW-260718 |
| PO no. | PO-4108 |
| Invoice date | 2026-07-18 |
| Due date | 2026-08-17 |
| Currency | USD |
| Total | 1480.00 |
A matching column name is not proof that a value is correct. Missing, ambiguous, or differently formatted source values remain a human-review decision.
Review the source, not just the field labels.
Use the actual signed-in workflow only after the output shape is clear. A well-named column cannot solve ambiguity in the source page.
Read the existing header row
Decide which column labels the downstream workbook, system, or reviewer actually needs. The product does not upload or modify the existing workbook.
Describe each field in plain language
Explain the source meaning, allowed format, missing-value rule, and ambiguity that must be reviewed. Keep fields in the destination column order.
Set the record boundary
For this public invoice, one record represents one synthetic invoice line. Real documents can repeat, omit, merge, or reorder information.
Review the source beside the output
Confirm field meaning, row boundaries, missing values, number formats, and ambiguous text before using a result operationally.
What this page does not establish.
It does not measure general extraction accuracy, speed, cost, or performance across scans, handwriting, multi-language documents, altered layouts, or missing source data. The invoice batch is a declared reference example, not a measured model result.
The interactive example does not upload a document, call a model, create a task, modify an existing workbook, store a template, or create a paid result. It only explains the field-to-column workflow.
Questions before you define the fields
What is AI document data extraction?
It is a workflow for requesting named fields or records from a document rather than simply keeping the full detected table. The result still requires source review.
Can I upload an existing Excel template on this page?
No. Use its header labels and business rules to define fields in the same order. FOVATA exports a new workbook; review it before copying or importing data into an existing system.
Is the interactive workbook example a live AI run?
No. The visible rows are declared values from public synthetic invoices, not live or historical model output.
Why show synthetic commercial invoices?
They provide public, non-personal sources that can be inspected beside declared reference workbooks. They do not prove performance for other document types.
When should I use the real extraction workflow?
Use the signed-in workflow when you have your own document and a defined field need. Review any result against its source before it informs a financial, customs, tax, or operational decision.
Start with the fields a reviewer can verify.
For an actual document, define the smallest useful set of fields, open the signed-in extraction workflow, and compare every result against the source before it moves downstream.
Open AI extraction