The scariest email I ever received as a finance manager was three lines long. Our biggest supplier, apparently, had changed banks, and would we kindly update the details before the next payment run. The logo was right, the tone was right, the timing matched a real invoice we owed. It was completely fake, and the only reason we didn't lose five figures that Friday is that our process forced a phone call before any bank detail change went live.
I've spent years building purchase to pay systems for UK businesses, first inside finance teams and then as a consultant, and that near miss shaped how I think about all of it. Automating vendor payments and purchase orders is genuinely one of the best projects a finance team can take on, but only if you're clear about which parts you're automating and which parts you're deliberately keeping human. One note on language before we start: the title says vendor payments because that's the phrase people type into Google, but in Britain we mostly say supplier payments, so I'll use the two interchangeably.
The Short Version
Automate the mechanical work and keep the judgement. Invoice capture, coding, chasing approvers, building payment files and reconciling the bank are all jobs software does faster and more accurately than a person. Deciding who can approve what, verifying a supplier's bank details, and pressing the final button on a payment run are jobs that stay with named humans, backed by rules the software enforces rather than replaces.
Done this way, automation increases control. Every purchase order, approval and payment carries a timestamped record of who did what, duplicate invoices get flagged before they're paid, and nobody needs your online banking password to get bills settled. The businesses that lose control aren't the ones that automated. They're the ones running approvals through email threads and payments through a shared bank login.
What Losing Control Actually Means
It helps to define the fear, because "losing control" gets used loosely. In accounts payable, the term for the money-out side of your books, control means three specific things. Money only leaves the account with the right sign-off. It only goes to genuine suppliers for goods or services you actually received. And you can prove both of those things afterwards, to an auditor, to HMRC, or to yourself at two in the morning.
The classic principle behind this is segregation of duties: the person who orders something shouldn't be the person who approves the invoice, and neither of them should be the person who releases the payment. In a manual setup this is hard to sustain, because one overworked bookkeeper often does all three. Good automation makes segregation the default, because the software physically won't let a requester approve their own spend.
So when someone tells me they're nervous that automation means "the system just pays things", I gently point out what they're comparing it to. A spreadsheet, a shared inbox, and a managing director approving invoices by replying "fine" from their phone. That's not control. That's habit.
What Manual Processing Really Costs
Let's put numbers on the boring version first. The most widely used benchmark comes from Ardent Partners, an American research firm, whose 2025 figures put the average all-in cost of processing a single invoice at 9.40 dollars, roughly £7, against 2.78 dollars for the best performers. The same research found the average invoice takes 9.2 days to get from arrival to payment, and that 14 percent of invoices become exceptions needing human intervention, against 9 percent for top teams. I should be honest that this is US research, because nobody produces a UK benchmark with the same rigour, but wage costs and processes are similar enough on both sides of the Atlantic that the direction of travel translates even if the pennies don't.
There is a UK-anchored figure worth knowing too. App Advisory Plus, which reviews accounting apps for UK practices, worked through the cost of manual approvals and found each approval request eats about an hour of someone's time once you include data entry and chasing, costed conservatively at the National Living Wage of £12.21 an hour. Automated workflows cut that to around twelve and a half minutes per request. Multiply either figure by a few hundred invoices a month and you can see why finance teams feel permanently behind.
And that's before the expensive mistakes. Duplicate payments, missed early settlement discounts, VAT reclaimed against invoices you can't later produce for HMRC. Making Tax Digital already requires VAT registered businesses to keep digital records, so every paper invoice in a lever arch file is a small compliance debt waiting to be called in.
The Fraud Numbers Every UK Finance Team Should Know
Here's the part that turns automation from a nice-to-have into a control project. According to UK Finance's Annual Fraud Report 2026, criminals stole £1.28 billion through payment fraud in 2025, and invoice and mandate scams accounted for £41.3 million of that across 2,305 cases. Do the division and the average invoice scam cost its victim just under £18,000. These are not lottery odds; they're targeted attacks on businesses exactly like yours.
An invoice or mandate scam is UK Finance's category for what happened in my opening story: a criminal intercepts or imitates a genuine supplier relationship and persuades your team to send a real payment to the wrong account. Reporting on the same UK Finance data by Market Business News notes that around £28 million of those 2025 losses fell on business and other non-personal accounts, and that only 48 percent of invoice and mandate scam losses were returned to victims. Less than half. Whatever the recovery process looks like, it is not a safety net you want to test.
The wider picture matters too. Authorised push payment fraud, known as APP fraud, is any scam where the victim is tricked into authorising the payment themselves, and it hit £576.4 million in 2025, up 19 percent on the year before. Your bank's security can't help you when your own team presses send.
Who Actually Gets Reimbursed
Since 7 October 2024, the UK has had mandatory reimbursement rules for APP fraud, and every business owner I speak to has half-heard about them and assumed they're covered. Most aren't. The Payment Systems Regulator, the body that oversees UK payment systems, requires banks to reimburse eligible victims up to £85,000 within five business days, with the cost split 50/50 between the sending and receiving bank.
The catch is the word eligible. As Tenet Law's analysis of the rules sets out, the scheme covers individuals, charities with income under £1 million, and microenterprises, defined as businesses with fewer than 10 people and turnover or balance sheet no more than £2 million. Employ your eleventh person or pass £2 million in turnover and you're outside the scheme entirely. The rules also only apply to Faster Payments and CHAPS transfers between UK accounts, claims must be made within 13 months, and your bank can apply a £100 excess and refuse claims involving gross negligence.
Read that back as a business owner and the conclusion writes itself. For most trading companies of any size, your internal payment controls are your fraud insurance, because the official scheme was designed for consumers and the very smallest firms. That's the frame I'd hold onto for the rest of this article.
Confirmation of Payee Is Free Control, So Use It
One control costs nothing and catches a remarkable amount. Confirmation of Payee, usually shortened to CoP, is the name checking service run by Pay.UK, the operator of the Faster Payments system. When you set up a new payee, the receiving bank checks whether the name you've typed matches the name on the account, and tells you it's a match, a close match, no match, or that the check couldn't be completed. More than 300 organisations now run CoP and over 2 million checks happen every day, which tells you how normal this has become.
My rule, and the one I write into every client's payment policy, is that anything short of a full match on a supplier payment is a stop, not a shrug. A close match on "Smith Building Ltd" versus "Smiths Building Limited" is probably fine, but you confirm it by phoning the supplier on a number you already hold, never one from the email that asked you to pay. CoP covers push payments over Faster Payments and CHAPS; it doesn't cover Bacs Direct Debits, so mandate changes need the same phone call discipline.
Good payment tools now bake this thinking in, which is one of the quiet arguments for automating the payment leg rather than typing account numbers into online banking by hand.
Build It in Four Layers, Not One Big System
The mistake I see most often is trying to buy one enormous platform on day one. Accounts payable automation is really four jobs, and it's fine, often better, to solve them with two or three connected tools. As the UK-focused comparison by Acenteus points out, most "best AP software" lists are US-built, priced in dollars and organised around card issuing and cheque runs that simply don't match how British firms pay suppliers, which is why names like BILL keep appearing in articles and rarely in UK finance stacks.
Layer one is capture: getting every invoice out of inboxes and photos and into structured data, using OCR, which is optical character recognition, software that reads documents so nobody retypes them. Layer two is commitment and approval: purchase orders raised before spend happens, and invoices routed to the right approvers automatically. Layer three is payment: building and releasing payment runs safely, with checks on bank details. Layer four is the ledger, your accounting software, where everything reconciles against the bank feed.
Your accounting platform anchors the stack. Xero is my usual anchor because I find its UK app ecosystem the deepest, but the same layered logic works on QuickBooks Online or Sage, and ApprovalMax, Dext and Lightyear all connect to more than one ledger, so you're choosing an architecture here, not a marriage. On that front, be aware that Xero's cheapest mainstream plan is a trap for payables work: as the AccountsOS breakdown of Xero's pricing notes, Ignite caps you at 10 bills a month, so most businesses automating supplier payments need at least Grow. Following Xero's price rise on 1 September 2026, the UK plans sit at £18 a month for Ignite, £39 for Grow, £55 for Comprehensive and £70 for Ultimate, all excluding VAT.
Purchase Orders and Three-Way Matching Without the Paper
Purchase orders have an image problem. People picture triplicate forms and procurement departments, so small firms skip them, and then wonder why invoices arrive that nobody remembers agreeing to. A purchase order, a PO, is simply approval before the money is committed rather than after it's spent, and that timing shift is where most of the control lives. Once a PO exists, the invoice isn't a decision anymore; it's a formality to be checked against something already agreed.
You don't need POs for everything. My usual advice is a threshold, commonly somewhere between £250 and £500, above which any commitment needs a PO raised and approved first, plus mandatory POs for stock, subcontractors and anything project related. Below the threshold, let approved invoices through on a simple workflow and save everyone's patience.
Matching is where automation earns its keep. Two-way matching compares the invoice to the PO. Three-way matching adds the delivery record, so you only pay for goods that actually arrived, which matters enormously in hospitality, retail, construction and anywhere with physical stock. Modern tools do this line by line: Lightyear, for instance, matches POs to deliveries and invoices, checks invoice line prices against your agreed supplier price lists to flag overcharging, and compares the bank details printed on an invoice against the details you hold on record. Clean matches under your threshold can then auto-approve, which is how the best teams get a decent chunk of invoices through with no human touch at all, while every exception still queues for a person.
The Tools I Rate for UK Businesses
Prices below are the published pound figures from the sources named, current at the time of writing and excluding VAT. A word of caution once, so I don't have to repeat it: some of these vendors have historically quoted plans in US dollars depending on where you look, so treat pound figures as close approximations until you have your own quote.
ApprovalMax. Best for: approval workflows and purchase orders on Xero, QuickBooks Online or NetSuite. It sits between your ledger and every payment, enforcing approval matrices, flagging duplicates and generating audit reports, and it's priced per organisation rather than per user, so you can add every budget holder without the bill moving. Software Advice's UK listing puts paid plans from about £35 a month, and reviewers there specifically praise being able to verify supplier bank details before they can be used. Note that ApprovalMax announced a move to tiered, usage based pricing from August 2026, sized to your documents and approvers, with its Capture OCR included on all plans and its open banking payment feature included for UK Xero customers, so check the current structure before you commit.
Dext. Best for: capture, receipts and expenses feeding your ledger. Independent UK analysis by ThriveOnz360 puts business plans at roughly £25 to £30 a month for five users and 250 documents, and fairly criticises Dext for not publishing a clean UK price list, so expect to start a trial to see real numbers. Watch the document cap. If you're on Xero already, Hubdoc is included in the subscription from Ignite upwards, as Receipt Router's plan guide confirms, and it's a perfectly serviceable free starting point for capture.
Lightyear. Best for: higher volume businesses with line-item complexity, think pubs, restaurant groups, retailers, builders. Capterra lists Lightyear from £130 a month for Starter, £279 for Business and £499 for Enterprise, on a usage based model with a 30 day trial, and Tekpon's review notes it's the Business tier that brings full three-way PO matching. It's not cheap, and I wouldn't put a 30 invoice a month consultancy on it, but for the right volume it replaces a part-time job.
Telleroo. Best for: safe bulk payment runs without handing anyone your bank access. You sync approved bills from Xero or upload a file, fund a ring-fenced account, and pay hundreds of suppliers in one go. Software Advice's UK listing shows paid plans from £49 a month, Telleroo's own pricing page explains the brackets scale with monthly payment volume, and the Capterra description highlights the bits I actually buy it for: alerts when a payee's bank details change, name matching against the account holder, and separate sub-accounts so the person paying suppliers never sees salaries.
Tipalti. Best for: mid-market and international payables, multiple entities, suppliers across currencies. Tipalti's own UK guide puts starter plans around £99 a month with custom pricing above that. Overkill for a typical SME, sensible once you're paying suppliers in six countries.
And one honest warning about what not to buy: nothing. The most overrated option in this market is doing it all inside a spreadsheet plus online banking because the software "costs money". You'll pay either way; the spreadsheet just invoices you in evenings and risk.
The Workflow I Set Up Most Often
For a UK company on Xero handling somewhere between 50 and 300 supplier invoices a month, here's the shape I deploy again and again, because it balances automation against control at every step. Invoices arrive at a dedicated bills address, never a person's inbox, and capture software reads them into draft bills with the VAT split checked. Anything over the PO threshold that lacks a matching purchase order bounces straight back to the requester, which trains the whole business faster than any policy document.
Approvals run through an approval matrix in ApprovalMax: budget holders sign off their own cost centres up to £1,000, the finance lead to £10,000, and the managing director above that, with nobody ever able to approve their own request. Matched, in-budget bills under a small threshold auto-approve. Twice a week, approved bills flow into a payment run in Telleroo, one person builds the run and a different person releases it, and any bank detail change or name mismatch stops that line until someone has phoned the supplier on the number we hold on file.
Reconciliation then happens almost by itself through the bank feed, and month end stops being archaeology. What I'd underline is how much of that paragraph is control language: thresholds, matrices, matches, two people, phone calls. Automating vendor payments and purchase orders properly is mostly the act of writing your controls down and letting software enforce them without fear or favour.
The Controls That Stay Human
Some things I never automate, and I'd be suspicious of any system that offered to. Changes to supplier bank details are verified by a call to a number you already hold, from your records or the supplier's official website, never from the email or invoice requesting the change, and the call gets logged. New suppliers get the same treatment before their first payment, along with a quick check that the company actually exists at Companies House.
Payment release always takes two people, however small the business, because dual authorisation is the single control that survives a compromised email account. Shared logins are banned, leavers lose access on their last day, and once a quarter someone senior reviews the approval matrix and the user list, because controls rot quietly when nobody owns them. None of this takes long. All of it is exactly what the fraud numbers earlier are pricing.
There's a record keeping payoff hiding in here too. HMRC expects VAT registered businesses to hold valid VAT invoices to support what they reclaim, and an automated system files every document against its transaction, retained and searchable for years without anyone thinking about it. The next VAT inspection becomes an export rather than an excavation, which is a sentence I never expected to write with genuine feeling.
What This Stack Costs Each Month, Honestly
For a small company the arithmetic is gentler than people expect. Xero Grow at £39, ApprovalMax from about £35 and Hubdoc thrown in free gets you capture, POs, enforced approvals and audit trails for roughly £75 a month excluding VAT. Add Telleroo from £49 when payment volume justifies it and you're near £125. A higher volume operation going the Lightyear route starts around £130 for the platform before the ledger and payments pieces.
Set those figures against one average invoice scam at just under £18,000, or the hour of staff time each manual approval burns, and the payback period is measured in weeks. I've never once had a client rip this out to go back to spreadsheets. I have had several arrive after a fraud, wishing they'd spent the £75 a year earlier.
The 2029 E-invoicing Mandate Changes the Timetable, Not the Logic
There's now a regulatory deadline sharpening all of this. At the Autumn Budget on 26 November 2025, the government confirmed that e-invoicing becomes mandatory for all VAT invoices in business to business and business to government transactions from 1 April 2029, as AccountingWEB reported at the time. An e-invoice here means structured, machine readable data flowing system to system, so a PDF attached to an email won't count, and HMRC has since confirmed Peppol, the international invoice exchange network, as the interoperability backbone, with detailed technical standards expected at Budget 2026.
Saffery's guidance for SMEs makes two points I'd echo. First, awareness is worryingly low, with roughly 40 percent of organisations unaware of the changes in HMRC's own research. Second, global research referenced by HMRC suggests a typical small organisation saves around £11,000 a year through e-invoicing, because structured data improves matching and cuts disputes. Helpfully, real-time reporting to HMRC is not part of the initial mandate, so this is about invoice format and plumbing rather than live tax surveillance, at least for now.
The practical takeaway is that everything in this article is the preparation. A business that already captures invoices digitally, matches them to purchase orders and pays through controlled runs will experience 2029 as a format change. A business still keying PDFs into a spreadsheet will experience it as a rebuild, on a deadline, alongside everyone else.
Where I'd Start
If you take one action from this, make it the bank details rule: write down, today, that no supplier bank detail change ever goes live without a call to a known number, and tell the whole team. It costs nothing and it defeats the single most expensive scam aimed at UK businesses. Then map who should approve what, pick the capture and approval layer that fits your ledger, and add controlled payments once the approvals are solid. Give yourself a fortnight per layer rather than attempting a big bang, and run the old process in parallel for one payment cycle so no supplier relationship is riding on day one.
Automating vendor payments and purchase orders isn't about trusting software with your money. It's about finally having a system strong enough to trust yourself with it, where the accounts payable automation does the drudgery, the purchase order approval workflow enforces your own rules, and every pound that leaves the account can explain itself. Start small, keep the humans where they matter, and let the machines chase the approvals for a change.