The function that brings money into your business is still run by hand.
Not mostly. Almost entirely. Whether you bill usage-based, seat-based, flat subscription, or some hybrid, the work of getting paid looks roughly the same: someone on your finance team is pulling data from multiple systems, building a billing schedule, sending invoices, chasing payment, reconciling what came in against what was owed, then doing it again next month.
73% of finance teams say the business is growing faster than they can keep up, according to Leapfin's 2025 State of Automation for Revenue Accounting survey of 200 finance professionals. That number was 48% a year ago. The gap between what the business needs and what the AR function can deliver is widening, not closing.
AP got Brex and Ramp. Payroll got Rippling and Deel. The functions that pay money out have been automated, audited, and largely solved. The function that brings money in is still done by hand.
That's not an oversight. AR stayed manual because it's structurally harder than the other two.
Why AR is harder than AP or payroll
AP is standardized on the input side. A vendor sends an invoice in roughly one of three formats. Software reads it, codes it to the right general ledger account, routes it for approval. If the number is wrong, the vendor calls eventually. The error is internal, the stakes are contained, and the approval policy fits on one page.
Payroll is deterministic. Federal tax tables, state withholding, FICA, benefits deductions. The rules are detailed but they don't change per employee. Build the rule engine once, run it at scale.
AR has none of those properties.
Every B2B contract is different. Volume tiers, usage ramps, true-ups, partial-period prorations, mid-term amendments. 35% of finance teams say new pricing structures, meaning hybrid and usage-based models, are the single biggest source of operational complexity they're dealing with right now. The output is customer-facing, which means an error doesn't sit quietly in your general ledger for six months. It shows up in your customer's inbox tomorrow morning. That's a support ticket, a relationship problem, and a churn signal in the same event.
Then there's ASC 606.
ASC 606, and its international equivalent IFRS 15, governs revenue recognition. Unlike the federal tax tables, it doesn't give you the answer. It gives you a framework for arriving at the answer, and that answer changes per contract. External auditors test it transaction by transaction.
An approximately right billing system is not a productivity tool. It's a Sarbanes-Oxley risk. SOX requires CFOs to personally certify their financial statements. A billing system that's approximately correct means a CFO personally signing off on an approximately correct revenue line.
The survey data reflects this. Only 43% of SaaS finance professionals say they are "very confident" their revenue recognition would hold up under audit or investor diligence. 7% of Controllers say they're not confident at all. Another 36% sit in the middle: "somewhat confident," which is another way of saying they know something is fragile but haven't found it yet.
There's also a data architecture problem. The contract lives in Salesforce. The pricing lives in your CPQ. The billing schedule lives in your billing system. The receivable lives in your AR ledger. The journal entry lives in your general ledger. Five systems, five slightly different versions of the truth, reconciled by hand every single period. When finance teams are asked which manual task they most want to eliminate, data reconciliation across systems comes first at 28%, followed by journal entries at 23%, and revenue recognition schedules at 17%. Those three tasks are the AR workflow, in order. AI that operates inside any one of those systems can only see one slice. That's a faster version of the existing problem, not a fix.
(Source: Leapfin State of Automation for Revenue Accounting, 2025, survey of 200 finance professionals across SaaS, e-commerce, and marketplace companies)
