Why AR got left behind
Accounts receivable automation has lagged AP by five to seven years in real adoption. Not because the technology wasn’t available but because the problem is genuinely harder.
AP has a clean shape. There’s a document (the invoice), a system that owns the truth (the ERP), and an unambiguous definition of done (payment posted). AR has none of that tidiness. Remittance data arrives in a dozen formats and rarely matches cleanly. Invoices get disputed. Collections is a relationship, not a transaction — the same customer you’re chasing this week is the one your sales team is trying to renew next week. That human dynamic is exactly why so many teams have been reluctant to point automation at it. It feels risky in a way that AP never did.
So AR stayed manual. And manual AR is expensive in a way that doesn’t show up on a software invoice. It shows up as a remittance file with 300 line items that someone matches by hand for three hours. As forty dunning emails sent one at a time from a shared inbox. As a collections rep piecing together a customer’s full picture across four systems before they can even pick up the phone. The work is enormous, it’s low-judgment, and it’s precisely the work that keeps the team from the high-judgment conversations that actually move a payment.
What agentic AR actually requires
Here’s the reframe that matters: the goal of AR automation is not to remove the human from collections. It’s to remove the human from the parts of collections that never needed one — so the team can spend its judgment where judgment is the whole point.
That takes three capabilities working together.
Remittance matching: Cash applied without keying. AI matches incoming remittance data to open invoices across formats, currencies, and partial payments. This is the single biggest low-judgment time sink in AR, and it’s the one that, once automated, frees the most human hours immediately. In Auditoria’s stack this is AR Remittances.
Dunning intelligence: Every account worked, every day, under your policy. Instead of a batch of reminders that fire on Monday and then go quiet, the system ranks receivables by likelihood to pay, generates personalized outreach within policy, and escalates the accounts that need a human with the full context already attached. This is AR Collections.
Dispute and inquiry triage: A consolidated answer, not a scavenger hunt. Inbound billing questions and disputes get classified, routed, and in standard cases resolved autonomously, with the customer’s whole account visible in one place rather than scattered across systems and inboxes. This is AR Helpdesk.
The through-line across all three is coverage at scale: every account worked, every day, under your policy — without adding headcount proportional to volume. That is what “DSO automation” actually means. Not a faster version of the old process. A different operating model for the half of the cash cycle that’s been running on human memory and a shared mailbox.
What it looks like in practice
Two companies make the point from different angles; one on what the work requires, one on what it returns.
Blackbaud: a SaaS company serving nonprofits, educational institutions, and healthcare, is a clean study in the requirements. Their business applications team described a collections operation living in and out of a shared Exchange mailbox, hopping across multiple systems to piece together what a customer actually owed. Their sharpest pain wasn’t volume; it was coherence. A customer with a broad mix of products might get a dunning notice for one invoice on Monday and another on Wednesday — no single, consolidated view of the account. Agentic AR gave them that consolidated view, and it smoothed the cadence: where dunning used to fire only on Mondays (a spike early in the week, then silence), work now queues up for the collections team every morning in a steady rhythm, and reps are assigned to accounts based on customer behavior rather than the luck of the batch. The requirement it satisfied wasn’t “send more emails.” It was “see the whole customer, and work every account evenly.”
Secureworks shows what that operating model returns. Facing a growing collections backlog across a large customer base, their team put agentic AR to work on exactly the three capabilities above and the numbers moved where it counts. DSO improved by roughly ten days: the business went from paying, on average, seven days late to five days early against terms. The collections team went from four people to two. Response times improved 67%, overall process workload dropped by about 80%, and AI now handles more than 90% of inbound AR inquiries. Same team’s attention, redirected from triage to the accounts that actually needed a human.
Requirements on one side, results on the other. The bridge between them is the operating model and not a tool bolted onto the old workflow.
The honest test
If you’ve automated AP and your DSO hasn’t moved, that’s not a failure of your AP project. It’s a sign you finished one leg of the cash cycle and left the other one manual. The good news is that the harder half is now tractable; remittance, dunning, and dispute work that genuinely required humans a few years ago can be covered at scale today, with the judgment cases escalated cleanly to the people best equipped to handle them.
That’s the difference between processing receivables faster and actually collecting them sooner. One is motion. The other is cash.
📎 Companion download: the one-page AR Automation Readiness Checklist five questions to test whether your cash cycle is actually automated, or just busy.