Touchless invoice processing: what it takes

Touchless invoice processing means a supplier invoice travels from the moment it arrives to a posted record in your ERP without a person handling it. Capture, extraction, matching, approval and posting all happen on their own for that invoice. Nobody opens it, keys a field, chases an approver or clicks post. The invoice arrives, and the next time anyone sees it, it is already a clean transaction in the ledger.

What touchless means

The word describes one invoice and its journey. A touchless invoice is one that no human touched between arrival and posting. Everything that would normally take an operator, reading the fields, coding the invoice to the right nominal and dimensions, matching it to a purchase order, routing it for sign-off, is done by the system without a pause for a person.

Your touchless rate is the share of invoices that make that journey on their own over a period. If a thousand invoices arrive in a month and 700 post without anyone touching them, that is a 70% touchless rate. The other 300 stopped somewhere and needed a human. Where they stopped, and why, is the whole subject.

A useful thing to notice up front is that touchless is not one feature you switch on. It is the result of several things working in sequence, and each one depends on the one before it.

The dependency chain

An invoice can only go touchless if every stage of its journey completes without a person. Those stages run in order, and a stage only matters once the one before it works. Get them out of order and you automate a problem you have yet to solve.

1. Data quality comes first. Before anything can be matched or posted, the fields have to come off the document correctly: supplier, invoice number, net, VAT, gross and the line items. If extraction is wrong, everything downstream inherits the error, and a confidently posted wrong invoice is worse than one that stopped for review. This is where reading the document accurately matters, and it is worth understanding the difference between IDP and OCR before you rely on either for straight-through posting.

2. Matching comes second. With clean fields in hand, the invoice has to reconcile against what you already know: the purchase order it belongs to and the goods receipt that confirms delivery. Three-way matching only works when the line-level data is reliable, which is why it depends on the first stage. A match that clears cleanly is what lets an invoice skip a human check.

3. Rules come third. Once an invoice is read and matched, your rules decide what happens next: the coding, the tolerances that allow a small price or quantity variance through, the approval thresholds, the posting logic for your ERP. Good rules only help once the data feeding them is clean and matched. Applied to messy input, they either block everything or wave through errors.

The chain is why teams that buy automation and stay stuck often have the order wrong. They tune approval rules while extraction is still shaky, or push for straight-through posting before their purchase orders are being raised properly. Fix the earliest broken link first. The stages above it cannot outperform it.

Why 100% touchless is the wrong target

It is tempting to treat touchless as a number to drive to 100%. That target does more harm than good, because exceptions exist for genuine reasons and most of them are reasons you want a person involved.

Some invoices should stop, every time:

• A price outside tolerance. The invoice bills more than the purchase order agreed. Someone needs to decide whether that is a supplier error, an agreed change or a problem.

• No purchase order. Spend that was never raised as a PO has nothing to match against. The control gap is the point, and posting it blindly defeats the purpose of matching at all.

• A new supplier with no history. The first invoices from a supplier are worth a human eye, both for extraction confidence and for fraud checks before bank details go on file.

• A suspected duplicate or a credit note. These carry real financial risk if handled wrongly, and judgement belongs with a person.

Push the touchless rate to 100% and you are forced to either weaken these controls or build brittle rules that guess at things a person should decide. The exceptions are where the value of a finance team sits. The goal is to send the clean, ordinary majority straight through so your team has the time to handle the genuine exceptions well.

Realistic rates and how to raise yours

There is no single benchmark for a healthy touchless rate, because it depends on how much of your spend runs on purchase orders, how clean your supplier data is, and which controls you require. Measure your own baseline first, then expect it to climb over time as the system learns your suppliers. The exact figure matters less than the direction and the reasons behind the invoices that still stop.

The formula is simple: invoices posted with no human touch, divided by all invoices processed in the period. Keep every invoice in the denominator, including credit notes and non-PO invoices, or the rate flatters you. Then list why each of the others stopped. The causes are where the improvement lives, as this example shows:

In month one, 610 of 1,000 invoices post untouched, a touchless rate of 61%. Two causes account for more than half of the exceptions: missing or wrong PO numbers, and prices outside tolerance. Asking the worst suppliers to quote the PO number and updating the agreed price lists halves both. Nothing else changes, and by month four 725 invoices post untouched, a rate of 72.5%. The remaining exceptions, such as goods not yet received, are the checks doing their job.

The share that can post on its own is anchored by how accurately the documents are read in the first place. Stratas runs at 89% field-level extraction accuracy, measured on real customer documents, and validation routes fields the model is unsure about, or that fail a check, to a person as exceptions before they post. Reading well is the floor everything else is built on.

To raise your rate, work on the inputs that decide it:

• Clean your supplier master data, so names, bank details and tax settings match and stop kicking invoices out for the wrong reasons.

• Get purchase orders raised before invoices arrive, so there is something to match against and PO-less spend shrinks.

• Set tolerances that reflect how your suppliers actually invoice, tight enough to catch errors and loose enough not to stop every rounding difference.

• Feed corrections back, so the fields a person fixes today read correctly next time and the same supplier stops causing exceptions.

Do that and the rate rises on its own, because you are removing the reasons invoices stop at the source. Forcing invoices through would only bury the same problems. Every invoice that goes touchless is one your team never had to open, and the ones that still stop are the ones that earned the attention.