Freehand's $75M: AI agents running procure-to-pay

Freehand raised a $75 million Series B to point autonomous agents at the least glamorous money in the enterprise: procurement, supplier onboarding, invoice checks, payments, and reconciliation. Not a copilot that drafts a PO for someone to approve — agents that own the procure-to-pay loop end to end. It's worth reading closely, because the leak they're selling against exists in every business that pays more than a dozen vendors.

What actually happened

Per Crunchbase News and the company announcement, Battery Ventures and NewRoad Capital Partners co-led the round on July 29, with Nexus Venture Partners and former Commerce Secretary Penny Pritzker participating. That takes the San Francisco company, founded in 2024, to $100 million raised total. Valuation isn't disclosed; CEO Nitin Jayakrishnan would only say it stepped up meaningfully from the $25M Series A. Customers named include Meta, Unilever, Johnson & Johnson, Pfizer, Cardinal Health, and Dunkin' — roughly 50 accounts, with payments moving across 60 to 70 countries.

The performance numbers are company-supplied and unaudited: 5–10% of spend recovered in complex categories, workflows finishing 5–7x faster, procure-to-pay cycle times down more than 70%. Treat those as the sales deck, not a benchmark. The structural claim underneath them is the interesting part — that this work is mostly document reading, tolerance checks, and follow-up email, which is exactly the shape of task agents handle without needing a frontier model.

Why procure-to-pay matters for your business

You have the same leak at 1/1000th the scale, and it's probably costing you more as a percentage. Duplicate invoices paid twice. A vendor who quietly raised rates 9% and nobody caught it against the contract. Early-pay discounts on net-30 terms you never take because approvals sit for three weeks. An annual renewal for software two ex-employees used. None of that requires an AI agent to find. It requires the data to be somewhere queryable.

So do it in that order. Pull your AP history — QuickBooks, Bill.com, your bank's export — into one table with vendor, date, amount, invoice number, and GL code. Run four queries: same vendor and amount inside 60 days, unit prices that moved more than 5% year over year, recurring charges with no owner, and average days from invoice receipt to payment. We've never run that pass for a client and found nothing. Then, once you know your own numbers, an agent has something to act against and a baseline that proves whether it worked.

Buy the automation second. The vendors raising $75M rounds are betting you'll pay them to fix a mess you could measure yourself in an afternoon.

Sources: Crunchbase News, GlobeNewswire.