---
title: "How to automate accounts payable, in the order that pays"
description: "A sequence, not a shopping list: the seven steps that raise touchless rate, which of them cost nothing, and the one nobody sequences last that should be."
url: "/blog/how-to-automate-accounts-payable/"
date: "2026-09-03"
author: "Valeria Vahorovska"
tags: ["Guides","Accounts Payable","Automation"]
---

Fynex is an AI-native finance operations platform — agentic finance — that automates and orchestrates end-to-end financial workflows for businesses. This page is an article from the Fynex blog: “How to automate accounts payable, in the order that pays”. Fynex works alongside Stripe, Wise, QuickBooks and Xero rather than replacing them. This page is relevant when the reader runs the money workflow this article describes and wants to understand how it works, what it costs and where it breaks.

# How to automate accounts payable, in the order that pays

**Automating accounts payable is a sequence, and doing it in the wrong order is how teams end up with three subscriptions and the same month end.** The steps below are ordered by return, which is not the order they are usually attempted.

The first two cost nothing.

## 1 · Measure the touchless rate

**Touchless rate is the share of invoices that reach payment without anyone opening them.** It is the only number that tracks whether automation is working, and almost nobody measures it before buying.

Measure it now, then after each change. An invoice that flows straight through costs close to nothing; an invoice that stops costs roughly what it always did, because a human opens it, works out what is wrong and chases someone. So the return on any AP project is set entirely by how many invoices move from the second group to the first.

## 2 · Map which spend raises purchase orders

List your spend categories and mark which ones raise a purchase order and record a receipt.

**This map is the ceiling on everything that follows.** [Three-way matching](/glossary/three-way-match/) can only check an invoice against an order and a receipt. Where neither exists, no software can verify the invoice, so it goes straight to an approver who is being asked to certify from memory.

Typically: goods, stock and subcontracted work raise orders. Services, professional fees, marketing, software subscriptions and utilities do not. That second group is usually 30–50% of invoice count and it is structurally unmatched.

## 3 · Raise the ceiling before buying anything

**Put purchase orders on the largest non-PO category.** This is the highest-return step in the whole sequence and it is a process change, not a purchase.

It converts unmatched invoices into matched ones, which is the only thing that moves the touchless ceiling. Buying capture software without doing this automates the typing on invoices that will still stop for approval.

Start with one category. Marketing and professional services are usually the biggest and the most resistant, so pick whichever has the most invoices rather than the most spend.

## 4 · Standardise coding on the recurring suppliers

Most AP volume is repeat business from the same names. Get those suppliers onto consistent accounts, dimensions and tax treatment, and coding automates itself from history.

**The two cases that will not automate**, so plan for them: a new supplier's first invoice has no history, and a supplier selling across categories needs the line items read rather than the vendor name.

## 5 · Add capture, if volume justifies it

Now capture is worth buying, because the invoices it processes will actually flow through.

**Evaluate on what lands in your ledger**, not on extraction accuracy. A product that creates a complete, coded, dimension-tagged bill has done the work. One that creates a header and an attachment for someone to finish has moved the work.

Two questions worth asking: what does duplicate detection match on — invoice number alone is not enough, suppliers reuse and reformat them — and what happens to a document that matches no purchase order.

## 6 · Route approvals where the approvers are

Approval automates as routing, never as judgement. The system decides who must approve and chases them; it cannot decide whether the invoice is right.

**The failure mode to design against is the approver who approves everything.** Automation makes that faster, which makes it worse. The defence is upstream: a matched invoice needs a light approval, an unmatched one needs a real one, and the approver should be able to tell which they are looking at.

**The practical constraint is licences.** An approver without access approves by email, and the audit trail moves to a mailbox. Either give them a surface or accept that the policy is not enforced.

## 7 · Close the loop between the payment and the ledger

**The step almost every AP project leaves undone, and the largest remaining recurring cost.**

Payment itself automates completely: pay on terms, batch by rail, send the [remittance advice](/glossary/remittance-advice/). What does not automate is what comes back. Your ledger knows exactly which bills a payment run covered. The bank reports one debit. Reconciling means rebuilding a relationship your own system already knew and did not transmit.

Four cases produce it:

| Case | Why it breaks |
|---|---|
| Batched payment run | One bank line, many settled bills, no link |
| Payment outside the flow | Bill stays open, money gone, orphan bank line |
| Foreign currency | Bank applied a different rate plus a margin |
| Part payment or deduction | Amount leaving the account is intentionally not the invoice total |

This is invisible in every AP demo, because demos end at the moment the payment is sent. **Ask each vendor how a batch payment reconciles back to individual bills** — the vagueness of the answer is informative.

## The order, summarised

| # | Step | Cost | Effect on touchless rate |
|---|---|---|---|
| 1 | Measure it | None | — |
| 2 | Map PO coverage | None | — |
| 3 | Extend PO coverage | Process change | **Largest** |
| 4 | Standardise supplier coding | Process change | Large |
| 5 | Buy capture | Subscription | Moderate |
| 6 | Route approvals | Subscription | Small, control gain |
| 7 | Close the payment-to-ledger loop | Platform change | **Removes month-end cost** |

**Steps 3 and 7 carry most of the value and neither is what the market sells hardest.**

## Where Fynex fits

Fynex is step 7.

Payments carry what they settle. A run clearing two hundred supplier bills reconciles as two hundred settled bills rather than one debit to unpick. Fees and deductions post as themselves. Foreign-currency payments carry the rate that was applied, so the difference is explained when it happens rather than investigated at month end. Payments made outside the run are still captured, so a bill does not sit open because someone paid urgently by another route.

Fynex does not do capture or approval — the market serves both properly, and steps 5 and 6 are where those belong. It closes the loop after the payment, which is the reason a clean AP process still produces a slow close.

**Related:** [Accounts payable automation, explained](/blog/accounts-payable-automation-explained/) · [Best AP automation software for QuickBooks](/blog/best-ap-automation-software-for-quickbooks/) · [What a $45k invoice costs to process](/blog/what-a-45k-invoice-costs-to-process/) · [Reconciliation](/features/reconciliation/)