Accounts payable document automation
docRefinery turns supplier invoices, purchase orders, credit notes and statements into checked, structured data through an API. Each supplier's layout is learned the first time it arrives and remembered after that, so a stable supplier list is the easiest workload there is. Fields it is unsure of go to a person, not to your ledger.
Updated
What it automates in accounts payable
docRefinery does the capture step of accounts payable: it turns supplier documents into checked, structured data. It reads a PDF, scan or photo of a supplier document and returns the fields your team would otherwise key in, with confidence scores and page citations. Approval, GL coding, matching and payment stay in your AP or ERP system.
What gets keyed from each supplier document
| Document | Fields teams need |
|---|---|
| Supplier invoice | Invoice number, supplier name and tax ID, invoice and due dates, PO reference, line items with quantities and unit prices, tax, total due, payment terms, bank details |
| Purchase order | PO number, buyer, supplier, order date, delivery address, line items, total |
| Credit note | Credit note number, the invoice it credits, reason, amounts and tax credited |
| Remittance advice | Payer, payment date, invoices settled, amount paid against each |
| Supplier statement | Statement date, open items, ageing buckets, balance due |
Invoices and purchase orders are in the 20-type catalogue, with field-by-field detail on the invoice and purchase order pages. Credit notes, remittances and statements are set up as a workflow or a custom type, trained at no charge.
Two hundred suppliers, two hundred templates
Accounts payable is the best case for docRefinery because your supplier list barely changes. A supplier's invoices come out of the same billing system on the same template every month, so a few hundred layouts can cover years of invoices. The first invoice from a supplier is learned once. Every invoice after it is remembered, not relearned. When a supplier redesigns its invoice, the new layout is re-trained without anything for your team to schedule (see vendor layout changes).
The advantage is only as large as the repetition. Two hundred suppliers billing monthly is two hundred layouts and thousands of remembered invoices. A long tail of one-off suppliers, each sending a single invoice a year, still gets extracted, but each one is new and gains nothing from memory.
Checks that catch invoice errors before posting
Workflow rules run on every document: line items add up to the subtotal, tax matches the printed rate, a PO reference is present. Each rule must pass its tests before it is saved. A failed rule flags the run for review instead of passing quietly. And when an invoice has no PO number, the field comes back null rather than a plausible guess.
Who signs off before an invoice posts
You set confidence thresholds per field, so total due and bank details can demand more than a line description. Anything under its threshold lands in a review queue beside the source page, with the value's location highlighted. Every correction is kept on the run. A correction can also become a suggested fix for that supplier's layout, replayed on that layout's past documents before anyone accepts it. docRefinery extracts the bank details that are printed. It does not check that an account belongs to the supplier, so a changed account still needs your payment controls.
What else an AP team gets
Layout memory and review are the headline, but most of the daily benefit comes from the parts around them.
- One format for five documents. Invoices, POs, credit notes, remittances and statements all return the same field and table shape, so matching code is written once.
- Field names your ERP already uses. A Return as alias on each field sets the output key, so
supplier_refornet_totallands in the right import column. - Line items as rows, with quantity, unit price and tax kept together for the match.
- New document types at no charge. Custom types are trained, hosted and re-trained for free, and a workflow starts from one sample document, with versions, replay and compare before it goes live.
- A view of where time goes. Insights tracks vendors, flagged fields and auto-approval trends, so you can see which suppliers cost reviewer time.
- Control over who does what. Governed fields, owner, admin and member roles, and invites keep the field set consistent across the team.
- An integration you can audit. API keys have an expiry, can be revoked, and keep an activity log.
- Overseas suppliers. 70+ languages and scripts are supported.
- Your own infrastructure if finance requires it: private VPC or fully air-gapped on premises, on Enterprise.
From capture to the ledger
Plugging in is mostly a one-off mapping. Name your fields once with Return as, point your code at the API (extract a document), and every document arrives in the same format. Without a developer, use a dashboard export: in Excel each table gets its own sheet, so line items stay together for matching, and corrected values are always included. CSV gives one row per invoice. The Chrome extension, in early access and not in the Chrome Web Store yet, auto-fills a web form such as a supplier portal. There is no native ERP connector or webhook today; both, along with pulling invoices from a shared inbox, are in design on the roadmap.
What manual invoice processing costs
Ardent Partners' Accounts Payable Metrics that Matter in 2026, published through Medius (an AP automation vendor that sponsors it), puts the average cost of processing a single invoice at $9.90, against $2.67 for best-in-class teams. It reports an average invoice exception rate of 19.9%, against 11.8% for best-in-class. These are third-party industry benchmarks, not docRefinery's measured performance.
AP teams that should look elsewhere
- Most of your invoices already arrive as structured e-invoices (XML or EDI). That data needs no extraction.
- Your suppliers are mostly one-off, so layouts rarely repeat.
- You want one product to do approvals, matching and payment. docRefinery is the capture layer underneath those.
- You need invoices pulled from a shared inbox automatically today.
Common questions
Does docRefinery approve or pay invoices?
No. docRefinery extracts invoice data (supplier, invoice number, dates, line items, tax, totals, bank details) and flags anything uncertain for review. Approval, matching, GL coding and payment stay in your AP or ERP system, which receives the data through the API or a CSV, Excel or JSON export.
Can docRefinery check that invoice totals add up?
Yes, with workflow rules. A rule can check that line items sum to the subtotal or that tax matches the printed rate, and it runs on every document. A rule must pass its tests before it is saved, and a failure flags the run for a person instead of letting the invoice through.
How much does accounts payable extraction cost?
The Starter plan is free for up to 100 pages a month. Paid plans are custom and quoted by sales on your volume, document types and deployment. See pricing.
Do credit notes and statements need a separate integration from invoices?
No. Every document type returns the same format: each field carries its value, your Return as alias, a confidence score and, where it can be placed, a location, and each table returns rows. Your code reads a credit note the way it reads an invoice, and only the field set differs. Credit notes are set up as a workflow or a custom type at no charge.
Sources
Related
- IntegrationsConnect docRefinery to your systems: REST API keys, workflows from one sample document, a Chrome extension, JSON, CSV and XLSX export, and on-premises options.
- InvoiceExtract data from invoice PDFs and scans: 62 fields including line items and VAT, as JSON with confidence scores and page and box citations on values.
- Purchase orderTurn customer and supplier purchase orders into order lines, terms and delivery dates. A PO OCR API with confidence scores and box citations on values.
- Invoice data extractionWhat invoice data extraction should return, why line items are the hard part, where it sits in accounts payable, and what to test in an invoice extraction API.
- Vendor layout changesWhy a sender changing their form breaks document extraction, how to detect a layout change early, how to handle it, and what to ask your extraction vendor.
- CompareHow docRefinery compares with Reducto, Extend, LlamaParse, Unstructured, Textract, Azure and Google on output, review and deployment, including where we trail.
- Create your account | docRefineryCreate a docRefinery account with Google. Starter is free up to 100 pages a month.
- Use casesWhere document extraction fits best: accounts payable, lending, KYC, insurance claims, logistics, healthcare admin and contracts, wherever senders repeat.
Bring the documents your team handles.
Upload a real file and see each field with its confidence score, ready to check.
100 pages a month free. No credit card, no sales call.