The Collections Desk That Scaled With the Book Was the Wrong One
Two lenders wrote the same volume. One added a collector for every increment of new accounts; the other held the desk flat and let follow-up scale on its own. The cost structures split.
Every alternative lender that grows its book eventually meets the same line item: back-office labor that rises in lockstep with volume. Collections is where it surfaces first, because more accounts mean more follow-up, and more follow-up has always meant more people.
Recent collection performance data challenges whether that tradeoff still holds. AI-run collection now recovers at two to four times the rate of traditional collection, at 60 to 80% lower cost per dollar recovered. Those figures do not describe an efficiency shaving. They describe a different cost structure, one where recovery capacity stops tracking headcount at all.
Why a manual desk scales linearly
A human collections desk has fixed daily capacity: a set number of calls, emails, and follow-ups per collector. When the book grows, coverage holds only if the desk grows with it. Lenders who staff to volume keep cost-to-collect roughly flat as a percentage, which reads as discipline until you notice recovery quality plateaus at the same time. Thirty collectors running a manual cadence cover the same proportion of the book that fifteen did. The extra headcount buys capacity, not reach.
That matters because the accounts a manual desk cannot reach are the expensive ones. A past-due account worked systematically from the first missed payment resolves in about 20 days, while a traditional agency works a roughly 180-day, six-month mandate for a materially lower share of the balance. And 15 to 20% of recoverable AR still exits through the coverage gap, the accounts no one had time to reach on cadence before they aged past the point of return.
The two lenders
Picture two lenders writing the same annual volume. The first treated collections as a staffing problem. As the book doubled, the desk doubled, cost-to-collect held flat as a percentage, and leadership called it managed. Recovery on the aging tail never improved, because a larger team running the same manual cadence covers the same slice of the book.
The second lender held the desk flat and put the volume on an operated cadence. Every past-due account entered a disciplined sequence from the first missed payment, coverage extended across the whole book rather than the top of it, and the labor line stopped moving with the book. Cost-to-collect fell as a percentage while recovery on the tail rose. Same volume, same market, and the only variable that differed was whether follow-up scaled with headcount or independently of it.
The gap compounds over time. The average business now carries 52 days of outstanding receivables, and roughly three-quarters of finance operations still run AR manually. Every one of those days is capital financing a customer instead of the lender’s own book, and a manual desk has no structural way to compress the number without adding people to it.
How CXO solves this
CXO’s Collections and AR Automation decouples recovery capacity from headcount. The system runs the full outbound sequence across the entire book at once: sequenced multi-channel outreach from the first missed payment, promise-to-pay and dispute states captured and logged to the CRM in real time, and escalation triggered by rule rather than by whoever has bandwidth that afternoon. Human effort concentrates on the exceptions that actually need judgment, a payment-plan negotiation or a disputed balance, instead of the volume of routine follow-up.
People do not leave the process. The process simply stops requiring a proportional number of them. A book that grows 40% no longer needs a collections desk that grows 40% to hold coverage, because the cadence that reaches every account is executed by the system and supervised by the team. Configured to the lender’s own rules, workflows, and systems, it operates inside the existing stack rather than replacing it, and it is built to run in production, not to sit as a pilot.
The decision in front of you
The question is not whether to automate collections in the abstract. It is narrower and more useful: when your book grows next year, does your collections cost grow with it by design, or by default? If the answer is that you would hire, model the alternative first. Take one segment of the aging book, measure what a disciplined full-coverage cadence recovers against what the current desk reaches, and price the difference in both recovered dollars and headcount avoided.
A lender that scales collections headcount with its book is not being cautious. It is locking in a cost line that rises with every new account and a recovery rate that plateaus regardless of how many people work it. The alternative is a follow-up engine that covers the whole book at a cost that no longer moves with volume, while the team spends its hours where judgment actually changes the outcome.
In most operations, far more work can be automated than leadership realizes. One discovery call is enough to size what automating it would return to your bottom line. Book it at https://cxocorporation.com/contact.