Multiline Invoice Recognition, Reinvented by AI
Some invoices are one line. Most aren’t. If you handle supplier invoices with dozens of line items spread across several pages, subcontractor charges by site, materials broken out by project, services split across departments, you already know that traditional OCR falls apart the moment a layout changes.
PaperLess’s multiline engine doesn’t work that way anymore, because it isn’t OCR wearing an AI label, it’s AI built to actually understand invoices, not just scan them. There’s no template to build, no zone to draw, no per-supplier setup before it can read a document correctly. Point it at a multi-line, multi-page invoice, however it’s laid out, and it reads it the way a person would: by recognising what a line item is, not by matching it against a pattern someone configured months ago. Reliably, near enough every time, regardless of how many pages it runs to or how the line items are arranged.
This is what AI was always supposed to do for accounts payable, not speed up manual work, replace the need for it. Once it’s captured every line, that same intelligence carries straight through into VAT reconciliation, nominal coding, and posting, without anyone re-keying a figure or nudging a line into the right cost centre by hand. It’s the most capable multiline reading technology PaperLess has ever built, and for accounts payable teams still treating complex invoices as the exception that needs manual handling, it’s the clearest sign yet of where AP is headed: full automation of invoice data entry, powered end to end by AI that actually understands what it’s reading.
Older invoice-capture tools ask you to teach them where to look, draw a box around the quantity column, mark the region where VAT sits, repeat for every supplier layout you deal with. Change suppliers, or have one supplier tweak their invoice template, and you’re back in setup mode.
PaperLess’s current multiline recognition drops that step entirely. The AI reads the structure of the page itself, what’s a line item, what’s a subtotal, what’s a summary row, the same way a person scanning the invoice would. It doesn’t matter whether the invoice is two pages or twenty, or whether the line items run to a handful or several hundred: accuracy holds steady. That’s a meaningful shift for construction firms, multi-site operators, and any accounts payable team that used to budget extra review time for their messiest suppliers.
It also means onboarding a new supplier for the first time is no different from processing one you’ve worked with for years. A subcontractor brought on for a single project, a one-off supplier used for an unusual purchase, a regular vendor who switches invoicing software without telling you, none of it requires a call to PaperLess, a support ticket, or a delay before that invoice can be processed. It arrives, it gets read, and it moves through your workflow the same as everything else.
This is the part that changes how AP teams actually work with the data, not just how it’s captured. Once an invoice is read, you can prompt PaperLess AI to present the information the way you actually need it, not just the way it happened to print on the page.
A few examples of what that looks like in practice:
Group costs by project or site. Got an invoice where charges for three different sites are interleaved across the line items? Ask PaperLess AI to group and total the costs by site, and it will, without you touching a spreadsheet.
Pull the summary, skip the detail. If all you need is the cost breakdown from the summary section of a long invoice rather than every individual line, say so. PaperLess AI reads it that way and hands back exactly that view.
Surface specific coding fields. Ask for the nominal codes, tax codes, departments, or project codes tied to particular line items, and PaperLess AI extracts and displays them, based on what’s actually shown on the invoice, no manual lookup, no guessing which line belongs to which cost centre.
In short: the invoice stays exactly as the supplier sent it. How you see the data afterward is up to you.
Multi-page invoices with line items scattered across departments, sites, or projects are exactly the documents this update was built for. A few common scenarios PaperLess now handles cleanly:
Construction and subcontractor invoices where costs need to be split and coded by site before they can be posted. Multi-department invoices, such as shared services or facilities charges spread across cost centres, that used to require manual re-keying. Long supplier invoices where finance only cares about the summary totals, not thirty individual line items. Invoices where the tax treatment or nominal coding varies line to line, and getting it wrong means a VAT mismatch down the line.
Multiline reading is only half the job, the numbers still need to add up. PaperLess continues to run automatic VAT reconciliation across every line, flagging mismatches before an invoice ever reaches posting, and the self-learning layer keeps applying nominal and VAT codes based on how your team has coded that supplier before. The AI reading update makes the front end faster and more accurate; the reconciliation and learning behind it hasn’t changed.
PaperLess connects directly to Sage (50, 200, Evolution, Intacct, Construction), SAP Business One, Xero, and Orderwise, so captured and coded invoice data posts straight through without a separate import step.