Lemkin says an agent cut late invoices from 56% to 8%
SaaStr handed quote-to-cash to an agent called 10K and says average days overdue fell from 17 to 6.
More than half of SaaStr's invoices were being paid late and nobody inside the company knew. Jason Lemkin says the real number only surfaced at month-end close, when 10K, the agent he calls SaaStr's AI VP of Revenue, reported that 56% of invoices were paid late while humans ran collections.
Along the way it found Bill reminders we’d paid for and never turned on, plus 2 customers still being billed for a product we stopped supporting 6 years ago.
Collections works well with agents because every input already sits in a system of record: the contract, Bill, QuickBooks. Very little judgment required.
After about a quarter of 10K running quote-to-cash, Lemkin reports late invoices at 8%, average days overdue down from 17 to 6, and the longest overdue invoice down from 51 days to 12.
What the agent actually does
The before picture explains a lot. Lemkin says part-time contractors without Salesforce access billed deals by hand a day or two a week, and follow-up was inconsistent.
Now, after every signature, the agent reads the full 20-page contract, because split payments and odd terms are scattered through it. It moves the deal to Closed Won in Salesforce, adds missing contacts, creates the invoice in bill dot com off the actual contract terms, and sends it within minutes to the signer, the deal contact, the business owner and a finance contact it finds through enrichment if the contract does not list one. Then it checks every open invoice daily, follows up weekly, and pulls a human in at seven days past due.
It also turned up some quiet money leaks. Lemkin says it found Bill reminders SaaStr had paid for and never switched on, plus two customers still being billed for a product the company stopped supporting six years ago.
The rollout was small
This was not a big-bang handover. Lemkin says 10K ran on three supervised deals before going autonomous on the fourth, with one wrong invoice since. After that it took over commissions in about 30 minutes.
His explanation for why this particular job worked is the useful part for anyone else trying it. Collections sits on top of systems of record that already hold every input the agent needs, the contract, Bill and QuickBooks, so very little judgment is required. Compare that to the jobs founders usually point agents at first, where the agent has to invent context nobody wrote down.
These are one company's numbers, self-reported, over roughly one quarter. Worth noting that Lemkin says the demand signal is real though. At SaaStr's Dreamforce booth, more people asked about agentic collections than anything else, and he reckons far fewer companies run it than run AI SDRs, even though almost every B2B company has the same problem.
Steal the numbers, not the vibes
One email a day. Under three minutes. Real MRR, real pricing, real launches from founders building with AI.