Invoice approval process that actually works
Most invoice approval processes are broken email chains. Here is how to build one with clear thresholds, three-way matching, and real accountability.
Invoice approvals sit at the intersection of procurement, receiving, and finance - and most companies treat them like an email forwarding exercise. Here’s how we think about structuring them at Tallyfy.
Accounts Payable Made Easy
Summary
- Three-way matching prevents fraud and overpayment - Comparing the purchase order, goods receipt, and supplier invoice before approving payment catches discrepancies that cost AP departments 0.8% to 2% of total disbursements in duplicate or erroneous payments
- PO invoices and non-PO invoices need different workflows - A PO invoice is pre-approved and just needs verification, while a non-PO invoice requires routing through multiple approvers because nobody agreed to the spend upfront
- Threshold-based routing keeps executives out of trivial decisions - A department manager can sign off on a $2,000 office supply order without dragging the CFO into it, but a $75,000 consulting engagement absolutely needs senior eyes
- The bottleneck is almost never the software - Manual invoice processing takes up to 20.8 days because approvals sit in someone’s inbox, not because the technology is slow. Fix the process first
Let me be blunt about something. Most AP teams I’ve talked to don’t have an approval problem. They have a routing problem. Invoices land in inboxes. Nobody knows whose turn it is. The person who should approve is on vacation. And by the time someone finally signs off, you’ve missed the early payment discount and annoyed a supplier who was counting on that cash.
This drives me crazy because the fix isn’t complicated. It’s just unsexy. You need clear rules about who approves what, at what dollar amount, and what happens when they don’t respond in time. That’s it. No magic.
Teams tell us the same thing in different words with workflow automation, the organizations that process invoices fastest aren’t the ones with the fanciest tools. They’re the ones who sat down once, defined their approval thresholds, and built escalation paths for when things stall.
What three-way matching is and why you can’t skip it
Three-way matching is the single most important control in accounts payable. You compare three documents before paying anything:
The purchase order - what you agreed to buy, at what price, in what quantity. This is the contract.
The goods receipt - proof that what was ordered showed up. Right items. Right quantities. Right condition.
The supplier invoice - the bill requesting payment for what was delivered.
When all three match, you pay. When they don’t, you investigate. Simple concept. Surprisingly hard to execute consistently.
Here’s why it matters. APQC research shows that bottom-performing AP departments need more than four times the staff to process the same volume - 14.4 FTEs per billion in revenue versus 3.3 for top performers. The difference? Top performers catch discrepancies before payment, not after. Three-way matching is how they do it.
Without it, you’re relying on someone in AP to remember that the purchase order said 500 units at $12 each but the invoice says 520 units at $13. Humans miss these things. Especially when they’re processing hundreds of invoices per week.
A two-way match (just PO and invoice, no receipt) is faster but riskier. You might pay for goods that never arrived. Or arrived damaged. Or arrived as the wrong thing entirely. For low-value purchases, maybe that’s acceptable. For anything material, three-way matching is non-negotiable.
PO invoices versus non-PO invoices
Not every invoice follows the same path. This is where a lot of AP teams trip up - they try to force every invoice through the same workflow when the two main types need fundamentally different handling.
PO invoices have a purchase order behind them. Someone already went through procurement, got approval, and issued a PO to the supplier. The hard work of authorization happened before the goods arrived. So when the invoice comes in, AP just needs to verify it matches the PO and the receipt. If everything lines up, pay it. No additional approval needed.
This is the efficient path. Pre-approved spend, verified delivery, straightforward payment. Medius reports that PO invoices are more transparent, less risky, and take significantly less time to process.
Non-PO invoices are a different beast entirely. No purchase order exists. Someone incurred an expense - maybe a one-off legal consultation, a software license renewal, an emergency repair - and now there’s a bill. Since nobody pre-approved this spend, the invoice needs to go through a full approval workflow before payment.
Non-PO invoices are where bottlenecks live. AP has to figure out the right GL code, find the right approver, get sign-off from budget owners, and sometimes escalate to senior leadership. All without the supporting paperwork that PO invoices provide.
Running Tallyfy taught us that the PO-to-non-PO ratio varies wildly across AP workflows. Some companies run 80% PO invoices and 20% non-PO. Others are the reverse. But the ones drowning in non-PO invoices are always the ones complaining about approval delays. The fix? Push more spend through the PO process upfront, even if it feels bureaucratic. The time you save on the back end is worth it.
Setting approval thresholds that don’t create gridlock
Here’s where I see companies mess up repeatedly. Either everyone needs CFO approval for everything (creating a massive bottleneck at the top) or there are no thresholds at all (creating zero accountability).
Neither works. You need tiered thresholds matched to roles.
A practical structure looks something like this:
Under $1,000 - Department manager approves. These are routine purchases - office supplies, small software subscriptions, minor repairs. The person closest to the spend should sign off. Don’t waste senior leadership time.
$1,000 to $10,000 - Department head or director. Bigger commitments that affect departmental budgets. The approver needs budget visibility but doesn’t need to be a VP.
$10,000 to $50,000 - VP or senior director. Material spend that could affect quarterly targets. At this level, you want someone with cross-departmental perspective.
Over $50,000 - CFO or finance committee. Major commitments that affect the company’s financial position. CASO’s research on approval thresholds confirms that invoices above $50,000 should require CFO authorization.
But thresholds alone aren’t enough. You also need:
Escalation rules. If the designated approver doesn’t respond within 48 hours, the invoice automatically escalates to their manager. Without this, invoices sit in limbo for weeks.
Delegation paths. When someone is on vacation, their approval authority needs to transfer to a designated backup. Not “whoever is around.” A specific person.
Split approval for sensitive categories. Some invoices need both budget owner approval AND finance approval regardless of amount. The engineering manager confirms the consulting work was delivered. Finance confirms the payment aligns with the contract terms.
What surprised us when we dug into the data that defining these rules once and encoding them into a workflow eliminates the “who do I send this to?” question permanently. That question, repeated hundreds of times a month, is where most of the delay comes from.
Bottlenecks nobody wants to admit
I’ve talked to dozens of finance teams about their invoice approval pain. The problems are remarkably consistent. And they’re almost never about software.
Missing information on invoices. The supplier sends an invoice without a PO number. Or with the wrong entity name. Or missing line-item detail. AP can’t process it, so they email the supplier. The supplier takes three days to respond. Meanwhile, the invoice sits.
This is a supplier onboarding problem disguised as an AP problem. If you set clear invoice requirements during vendor setup - required fields, format expectations, where to send invoices - you eliminate this at the source.
Approvers who don’t approve. Stampli’s research shows manual invoice processing takes up to 20.8 days. The invoice itself takes minutes to review. The other 20 days? Sitting in someone’s inbox.
The fix isn’t nagging people. It’s automatic escalation after a defined window. If John doesn’t approve within 48 hours, his manager gets notified. Not as a CC. As the new approver.
No separation between verification and approval. Verification asks: does this invoice match what we ordered and received? Approval asks: should we pay this? These are different questions answered by different people. When one person does both, mistakes compound.
Duplicate invoices. Same invoice, different format. The supplier emails a PDF, then mails a paper copy, then follows up with another email. Without systematic matching, AP might pay it twice. Research shows 1.29% of processed invoices are duplicates, averaging $2,034 each. On a thousand invoices a month, that’s real money walking out the door.
Why AI doesn’t fix this (yet)
I know that sounds like a bumper sticker, but I’ve watched it play out. A company with a chaotic invoice approval process bolts on an AI tool that reads invoices and extracts data. Great. Now they can extract data from invoices faster. But the invoices still sit waiting for approval because nobody defined who approves what. The bottleneck didn’t move. It just got a fancier front door.
Parseur’s benchmarks show AI can extract invoice data with high accuracy. But extraction isn’t the problem for most companies. Routing and approval is the problem.
Here’s where AI does help: after you’ve defined your process. Once you have clear thresholds, escalation rules, and matching criteria, AI can flag anomalies. A supplier who usually invoices $5,000 a month suddenly submits $50,000? Flag it. An invoice arrives for a PO that’s already been fully paid? Block it. Line items that don’t match the receipt quantities? Route it to exception handling.
But all of that requires the process to exist first. The AI needs something to operate on. Without defined workflows, it’s just a very expensive data entry tool.
This is the problem Tallyfy was designed to solve - process definition comes first, automation comes second. You can’t automate what you haven’t defined.
Building the workflow from scratch
If you’re starting fresh or replacing a broken process, here’s the sequence that works. I’ve seen variations of this across dozens of implementations.
Map your invoice types. List every category of invoice your company receives. PO-backed purchases. Recurring subscriptions. One-time services. Expense reimbursements. Utilities. Each needs its own path, even if some paths are similar.
Define your matching requirements. Three-way match for PO invoices above your threshold. Two-way match for low-risk, low-value items. Receipt-optional for recurring fixed-cost services like rent or insurance where the amount doesn’t change.
Set your thresholds. Use the tiered structure above as a starting point, then adjust based on your company’s risk tolerance and organizational structure. A 50-person company might collapse everything into two tiers. A 5,000-person company might need five or six.
Build exception handling. What happens when the invoice doesn’t match? Who investigates? What’s the resolution timeline? How do you communicate back to the supplier? This is the part most companies skip, and it’s the part that causes 80% of the delays.
Add escalation timers. Every approval step needs a clock. Forty-eight hours is reasonable for routine invoices. Twenty-four hours for urgent or time-sensitive payments. The clock should trigger automatic escalation, not just a reminder email that gets ignored.
Test with real invoices. Take your last 50 invoices and run them through the new workflow on paper. Where do they get stuck? Which thresholds feel wrong? Which approvers get too many or too few? Adjust before you go live.
The biggest lesson from our own journey is that companies who spend two weeks designing the workflow properly save months of headaches after launch. The ones who rush into automation without doing this design work end up rebuilding everything within six months.
What good looks like in practice
APQC benchmarks draw a stark line between top and bottom performers. Top-performing AP departments complete the full cycle from invoice receipt to payment in 2.8 days or less. Bottom performers take a week or longer. The median sits around four days for invoice receipt to approval.
The gap isn’t about technology. Both groups have access to the same tools. The difference is process discipline.
Top performers share a few traits:
They route invoices to the right approver immediately, without manual triage. They enforce approval deadlines with automatic escalation. Discrepancies get caught during matching, before the approval stage, not after. They separate PO invoices from non-PO invoices at intake and process them through different workflows. And they measure everything - cycle time, exception rate, cost per invoice.
CFO.com reports the cost spread is just as dramatic. Top performers process invoices for roughly $1.77 each. Bottom performers spend $10.89. The median is somewhere around $9.87. That difference on ten thousand invoices per year is over $80,000 - enough to fund the process improvement project that would fix the problem.
Honestly, the math is so obvious it’s embarrassing. But most companies never measure their cost per invoice, so they never see the opportunity. They just keep doing what they’ve always done, manually routing emails and hoping for the best.
The invoice approval process isn’t glamorous. Nobody is going to make a keynote speech about it. But getting it right - clear thresholds, proper matching, automatic escalation, separate paths for PO and non-PO invoices - is one of those boring operational wins that compounds quietly. Fix it once, and it stays fixed.
About the Author
Amit is the CEO of Tallyfy. He is a workflow expert and specializes in process automation and the next generation of business process management in the post-flowchart age. He has decades of consulting experience in task and workflow automation, continuous improvement (all the flavors) and AI-driven workflows for small and large companies. Amit did a Computer Science degree at the University of Bath and moved from the UK to St. Louis, MO in 2014. He loves watching American robins and their nesting behaviors!
Follow Amit on his website, LinkedIn, Facebook, Reddit, X (Twitter) or YouTube.
Automate your workflows with Tallyfy
Stop chasing status updates. Track and automate your processes in one place.