Building a Scalable AR Function Without Adding Headcount

AR teams don't have a hiring problem. They have a friction problem, and it's an expensive one: $1.7 trillion sits trapped in excess working capital across the top 1,000 U.S. public companies, with receivables accounting for roughly $600 billion of that gap. Adding another collector doesn't fix whatever's actually slowing every invoice down; it just puts one more person in front of the same broken process, waiting.

How much working capital is sitting in underperforming AR right now

The Hackett Group's 2025 U.S. Working Capital Survey put a number on something finance leaders have felt for a while. Gross working capital at the top 1,000 U.S. public nonfinancial companies sits at 35% idle, roughly $1.7 trillion, when it could be out working instead of parked.

AR is the biggest single piece of that pile, and Hackett pegs it around $600 billion, driven by an 18-day DSO gap between top-quartile companies and the median. Companies got better at the cash conversion cycle overall in 2025, down 4% to 37 days, but nearly all of that gain came from payables. Receivables and inventory both got slightly worse, since somebody paid their bills later on purpose while collecting slower by accident, and nobody planned it that way.

The Hackett Group's research points to working capital optimization as a leading priority for finance leaders, with process changes, skill development, and technology consistently highlighted as the levers that matter. That distinction matters more than it sounds: the 18-day gap has more to do with what happens to an invoice after it leaves the building than with which customers are creditworthy. Close even part of that gap and cash stops sitting in somebody's inbox, waiting on a reply that isn't coming, and starts sitting in the bank instead.

What manual AR actually costs per invoice and per employee

Finance professionals spend up to 20 hours a week on manual invoice processing, which amounts to half a job, gone, before anyone's even started collecting a dollar.

A solid AR specialist handles somewhere between 200 and 350 invoices a month before the work starts slipping. Push past that range and things get missed, quietly, until they don't. Teams processing under $1 million in receivables per FTE each month carry a collections job title while doing work that barely resembles collections at all. Mostly it's clerical: chasing down a PO number, re-sending a PDF, checking a portal for the third time that week.

The dollar figures are blunt. Manual invoice processing runs $10 to $15 per invoice; automated processing runs $2 to $3, a five-times difference, repeated thousands of times a month. Manual processes also tend to add 15 to 30 days to DSO, so money that could be working sits parked, earning nothing.

The fix everyone reaches for first, hiring another collector, costs $65,000 to $85,000 a year per person. That expense repeats at every growth stage while the underlying slowness never improves; it just gets more expensive to ignore. PYMNTS Intelligence found in 2024 that 59% of U.S. businesses blame poor cash flow forecasting on outdated manual AR methods. Doing nothing shows up as a recurring salary line, keeping a process running that's exactly as clunky at $50 million in revenue as it was at $10 million.

Why automation alone hasn't solved the problem for most teams

Automation was supposed to fix this, but it hasn't. The numbers say why: PYMNTS found 83% of firms still haven't fully automated AR, and PwC's 2024 Finance Effectiveness Benchmarking Report found that even top-quartile companies have only automated 8% of AR tasks. Most of the work at the best-run companies in the country still gets done by hand.

Here's what most people get backwards: they assume automation failed because companies didn't buy enough of it. That diagnosis misses the real pattern. Partial automation grabs the easy stuff and leaves the hard stuff exactly where it was standing. A bot built to log into a customer portal and pull a payment status breaks the moment that portal redesigns its interface, which happens more often than anyone budgets for. Automated dunning emails go out on schedule and get ignored, because a templated message doesn't adapt to whatever's actually going on with that customer's business this month. Systems flag exceptions, sure, but flagging just leaves resolution, the slow, judgment-heavy part, sitting on somebody's desk.

Most teams still steer by DSO and basic aging reports, ignoring the Collection Effectiveness Index, bad debt ratios, and FTE productivity numbers that would actually show them where the time disappears. Automation got pointed at moving data around: sending invoices, matching payments. The judgment calls, sequencing follow-ups, handling a dispute, navigating ten different supplier portals with ten different login screens, stayed manual. That's why headcount keeps climbing at companies that will tell you, with a straight face, they've "automated AR."

The specific operational friction points that make each invoice expensive

Supplier portals are a good place to start, mostly because they're everyone's least favorite place to start. Coupa, Ariba, and their cousins each require a manual login, a manual upload, manual status checks. Each has its own interface, its own document requirements, its own exception process. A team managing a few dozen enterprise accounts might juggle a dozen different portal logins with no standard way to work through any of them. Call it a scavenger hunt with a deadline attached, except the prize is just getting paid for work already done.

Missing documentation is its own tax. A W-9, a certificate of insurance, a PO number, a remittance detail: some customer somewhere always needs one more thing before releasing payment. These requests show up mid-cycle, right when an invoice looked like it was finally about to clear, and reset the clock along with the collector's patience.

Then there's plain old communication breakdown. Follow-up emails land in the inbox of someone who left the company eight months ago, or a customer replies to confirm they got the invoice, then goes silent on when they'll actually pay it. Some situations need a phone call, a different tone, a different person entirely, and that shift doesn't happen on its own.

Disputes are the worst of the bunch. One disputed line item freezes an entire invoice, even when the disputed amount is a sliver of the total owed. Untangling it means pulling source documents, looping in another department, resubmitting, all of it by hand, all of it slow.

None of this shows up in a DSO report until weeks later, once the damage is already sitting on the books. That's exactly why most teams never trace a bad DSO number back to the portal login that took four tries and a password reset.

What a scalable AR operating model actually looks like

Buying software alone doesn't scale AR; it's the first, incomplete step. What actually scales is an operating model built in three layers, and skipping any one of them is why so many "automated" teams still feel manual.

Layer one automates everything that doesn't need a human brain attached to it: invoice delivery, payment matching, remittance capture. None of that requires judgment, so none of it should require a person clicking through it.

Layer two uses AI to sort what's left: which accounts to call today, which disputes need escalation, which customers are drifting toward 90-plus days. A prioritized queue beats a chronological list every time, because not every overdue invoice is equally urgent or equally collectible. Treating them the same wastes the hours that actually move cash.

Layer three is follow-up that behaves like a person paying attention, tuned to what actually happened rather than the same email going out on day 30, 45, and 60 regardless of what the customer said last time. Outreach should respond to the conversation already had, not run on a fixed script that ignores it.

Cash application is the floor everything else stands on. AI-native platforms can hit straight-through processing rates above 95%, compared to the 60 to 75% typical of manual matching. Billtrust's 2026 AR Benchmark Report found 92% of payments now need zero manual intervention, up from just over 90% in 2024. On top of that foundation sits the collections intelligence: worklists ranked by who's actually likely to pay rather than whose invoice is oldest, agentic AI that logs into a portal and submits documentation on its own for routine cases, escalation logic that saves a senior collector's attention for disputes that genuinely need judgment. Cloud platforms held 79.21% of the AR automation market in 2025, growing at a 12.11% compound annual rate through 2031, because updates need to roll out without downtime while payment rails and invoice formats keep shifting underneath everyone.

What finance teams should measure to know whether the model is working

DSO tells you something, but not enough. A DSO under 45 days is the commonly cited healthy mark, and it still hides exactly where in the process the time gets lost.

Collection Effectiveness Index above 80% is a more honest number. It shows whether the team actually converts collectible receivables into cash, and it catches a backlog quietly building even while DSO looks fine on paper. Pair that with bad debt ratio: under 1.5% is the bar for high performers heading into 2026, per Esker's analysis of AR benchmarks. That figure is really just the downstream scorecard for how well the upstream friction got handled.

FTE productivity matters too. Teams processing under $1 million in receivables per FTE each month have room to fix their process before they hire anyone else. The 18-day DSO gap between top-quartile and median performers is a useful outside yardstick, but the more useful number is internal: the gap between current DSO and what the customer base would actually support if the process ran with no friction at all.

What most teams skip entirely: portal submission success rates, first-contact resolution on disputes, and the share of invoices needing more than two follow-up touches. These are leading indicators, and they show where trouble is forming before it shows up in next month's aging report. Most teams manage exclusively to lagging numbers, and that gap, between watching what's about to happen and watching what already did, is a big part of what separates a scalable AR function from one that's just getting by.

What realistic improvement looks like and how quickly it arrives

The range is wide, so start there instead of pretending there's one magic number. Companies that digitize AR commonly report meaningful DSO reductions. Some documented deployments show 33-day DSO reductions alongside significant improvements in aging balances.

A couple of real examples make the range less abstract. Some deployments have shown dramatic DSO compression within months of going live. Others have reported more modest but still material reductions in the range of days, not weeks.

Speed matters as much as size. Agentic, AI-native platforms have shown meaningful DSO reductions early after going live, faster than most finance transformation projects ever pay back. Run the math on a $200 million revenue company: even a modest DSO improvement frees up millions in working capital. That's the number that should reframe how any AR investment gets judged internally, ahead of any vague promise of "efficiency."

For finance leaders staring down a hiring decision, the takeaway is plain, and it's not a close call. A manual AR team has a productivity ceiling that doesn't move no matter who's standing under it. A well-built AR operating model raises its ceiling as volume grows, without a matching rise in headcount. The one honest caveat: results depend entirely on which friction points actually get fixed. Automate invoice delivery while leaving portal logins and follow-up sequencing manual, and the gains show up half-finished, like a renovation that redid the kitchen and left the plumbing alone. The real payoff needs all three layers working together, not one bolted onto an otherwise unchanged process.