How It Works: One Mixed Batch, Start to Sorted

Stage 1

The batch arrives: mixed, unnamed, unsorted

The batch lands the way batches always land: a dozen supplier invoices in one email, two claim forms scanned slightly crooked, a signed contract as a phone photo, a credit note named "scan_final2.pdf," statements, correspondence, and one attachment nobody can identify from its name. Nothing is pre-sorted, and nothing needs to be.

Arrival is intake. Every document is captured the moment it lands, given an identity, and queued for the read — email attachments, uploads, scans, and system feeds all enter the same way.

Stage 2

Every document is identified — by what it is

The classifier — MaxxOwl™ SI on the MaxxLogix platform — reads each document in turn. It decides from the document's layout, its language, and its content: the shape of an invoice table, the wording of a claim, the clauses and signature blocks of a contract. "scan_final2.pdf" is a credit note, and it is labelled as one.

File names, folders, and senders are never consulted. The read stands on the document itself, so a sender's habits can't misroute it.

Stage 3

Data is lifted — with a confidence level on every value

Classification finished, extraction follows immediately, because knowing the type means knowing which fields matter. From the invoices: vendor, invoice number, dates, line items, tax, totals. From the claims: claimant, reference, dates, amounts. From the contract: parties, term, renewal date.

Every value arrives as structured data with a confidence level attached. A crisp total reads with high confidence. A smudged figure on a crooked scan does not — and the read says so.

Stage 4

The unsure go to people — briefly

Anything below the confidence bar leaves the straight-through path and lands in Outcome Workspace™, where a person sees the document and its extracted data side by side: here is the scan, here is what was read, here is the value in doubt. The person confirms or corrects it in seconds.

The decision is recorded with the document. Human judgement is spent exactly where the read asked for it — and nowhere else.

Stage 5

Sorted, routed, and on the record

Now the batch disperses into the work it represents. Invoices move to matching against their purchase orders and receipts. Claims enter their review workflow. The contract is filed to the Outcome Repository™ with its renewal date as data. Correspondence is routed to its owner.

And the read itself is on the record: volumes by type, confidence and exception rates, and the outcomes each document went on to produce — what Outcome Analytics™ reports on, and what got done with every document in the batch.

Nothing about the batch is left to memory. If anyone later asks why a document went where it went, the answer is in the record: what it was classified as, what was extracted, which values a person verified, and what happened next.