Fazeshift

Here is a thing about companies: they sell stuff, and then they would like to be paid for the stuff, and the second part turns out to be shockingly hard. Not hard in the sense of physics. Hard in the sense that the money is owed, everyone agrees it is owed, and yet getting it into the bank account requires a person to open NetSuite, then Salesforce, then a bank portal, then their email, and manually reconcile numbers that live in four systems that refuse to speak to one another.

Fazeshift is a company that looked at this and concluded, reasonably, that a software agent should do the bouncing between tabs instead of a human. That is the whole pitch, and it is a better pitch than it sounds, because "getting paid" is not a niche. Every business that invoices has this problem. There are, by Fazeshift's count, more than a million accounts receivable clerks in the United States alone, and a large share of their day is copying figures from one screen to another.

The founders, Caitlin Leksana and Timmy Galvin, did not arrive at accounts receivable because it was glamorous. They arrived because it hurt. At a previous startup called Carma, they found themselves color-coding a spreadsheet to track payments from ten - ten - customers. If that is unbearable at ten customers, it is a genuine crisis at ten thousand, and crises at scale are where companies come from.

What Fazeshift built is less a product you log into and more a layer that sits on top of the tools you already have. The AI agents generate invoices, match incoming payments to the right open invoice, chase the ones that are late, write the reminder emails, escalate the awkward ones, and reconcile the cash - across ERPs, CRMs, payment platforms and inboxes. The company describes it as an "intelligent control layer." Across its customer base, it says the agents now handle more than 90% of the manual work.

The interesting strategic claim is not "we automate tasks." Lots of software automates tasks. The claim is that Fazeshift gets smarter with every invoice, because it accumulates proprietary data on how each payer actually behaves - who pays on time, who needs three nudges, who always disputes line item four. That is the kind of advantage that compounds, and compounding advantages are the ones investors pay for. Which, in May 2026, they did: $17 million in a Series A led by F-Prime, bringing the total to $22 million.

The people writing checks are not naive about the category. Gradient, Google's dedicated AI fund, led the seed and came back for the A. Y Combinator, where Fazeshift launched in the Summer 2024 batch, participated too. The revenue grew 12x in a year, which is the sort of number that makes a boring category suddenly very interesting to everybody.