Invoices · receipts · contracts · CVs
Stop retyping numbers off a PDF. And never trust one that was guessed.
Drop in a document, name the fields you want, and get them back as data. The part that matters is what happens when a field isn't there: it comes back empty with a reason, not filled in with something plausible. Every value that is returned quotes the exact text it was read from, so you can check it in seconds instead of re-reading the document.
A wrong total that looks confident is worse than no tool at all, because nobody double-checks a number that arrives neatly formatted in a table.
| Field | Value, and where it came from | Confidence |
|---|
Line items
| Description | Qty | Unit price | Amount |
|---|
What this is. A demonstration of a document-reading pipeline, not a bookkeeping system. It reads one document at a time and keeps nothing. Check anything marked medium or low before it goes into an accounts system — the confidence and the quoted evidence exist so that checking takes seconds rather than a re-read.
What this is actually for
The demo reads one document. The version worth paying for reads the inbox: every supplier invoice that arrives, matched against the purchase order, routed for approval, written into the ledger. This page is the engine underneath that, working, so you can judge it before commissioning anything.
Accounts payable
Supplier invoices into your ledger without a person retyping them, with the exceptions flagged rather than silently wrong.
Contract intake
Renewal dates, notice periods and liability caps pulled out of signed PDFs into something you can actually search.
Recruitment intake
CVs into structured candidate records, with the source text attached so a hiring decision never rests on a paraphrase.
Expense processing
Photographed receipts into claim lines, with totals that reconcile or say why they do not.
If one of those is a real cost in your business, tell me what the document flow looks like and I will tell you honestly whether it is worth automating.