On July 29, Freehand raised a $75M Series B co-led by Battery Ventures and NewRoad Capital Partners, with Nexus Venture Partners and Penny Pritzker's PSP Growth participating. The money runs autonomous "AI Teams" that manage supply-chain spend for the Fortune 500 — negotiating rates, enforcing contracts, processing payments, and deciding which invoices get paid — already deployed at Meta, Unilever, Johnson & Johnson, and Pfizer. Total funding is now $100M; the round is a Series B for a company founded in February 2024.
The number is not the story. A $75M vertical-agent round is on-trend, not a surprise — July's funding wave already split into "control the agents" and "own a regulated vertical", and this is firmly the second bet. The story is who raised it.
The one-line read: Freehand's founders already built the SaaS version of this exact workflow, sold it, and came back to rebuild it as agents. That's not a pivot into AI — it's a considered claim that the same domain is worth more as a workforce than as software.
The tell is the founders, not the round#
Nitin Jayakrishnan and Abhijeet Manohar previously built Pando, a procure-to-pay and transportation-management SaaS for large enterprise logistics customers. They sold it to a strategic buyer in early 2026 — a clean, full shareholder exit. Then they started over, in the same domain, with the same buyers, and built agents instead of dashboards.
Read that sequence carefully, because it's the most legible version of a bet the whole sector is making. These are not model-lab refugees chasing a demo. They are operators who shipped the software incumbent, watched exactly where it left value on the table, and concluded the leftover value was the part software couldn't do: the judgment calls a procurement team makes on top of the system of record. Freehand is funded to automate the team, not the tool.
What "own the vertical" actually buys you#
The horizontal pitch — a general agent platform anyone can point at any workflow — is a race to whoever has the cheapest tokens and the best harness. The vertical pitch is the opposite: a narrow, unglamorous, rule-bound workflow where the moat is domain knowledge and the enterprise's permission to act on money.
| Pando (the SaaS they sold) | Freehand (the agents they built) | |
|---|---|---|
| What you buy | A system of record + workflow tooling | A workforce that executes the workflow |
| Who does the work | Your team, using the software | Autonomous AI Teams; humans on exceptions |
| Unit of value | Seats and modules | Outcomes — spend recovered, cycles closed |
| Where the moat sits | Integrations + data model | Domain judgment + permission to act |
Freehand reports customers recovering 5–10% of spend, closing workflows 5–7x faster, and cutting procure-to-pay cycle times by more than 70%. Treat those as company-reported until an enterprise says them out loud — but note the shape: every figure is an outcome, not a feature. That is the vertical-agent business model in one line. You don't sell software that might help; you sell a result and take responsibility for it.
The hard part isn't intelligence. It's the signature.#
An agent that decides which invoices Pfizer pays is not a model problem. Any frontier model can rank invoices. The problem is that a Fortune 500 CFO has to be willing to let it — and to sign the audit trail afterward. That's why this is a regulated-vertical bet, not just a vertical one: the same governance pressure that's reshaping who can ship an AI product is exactly what makes a trusted spend agent defensible. If getting an agent cleared to move money is hard, then being the company that already cleared it at Meta and Unilever is the moat.
For founders, three things fall out of this deal:
- Domain beats model. If you deeply understand a spend-heavy, contract-bound enterprise workflow, that knowledge — plus the trust to act on it — is more durable than any model you could fine-tune. The lab ships the intelligence; you ship the judgment and the accountability.
- Sell the outcome, own the risk. Vertical agents that price on results have to eat the downside when the agent is wrong. That's a feature, not a bug: it's the thing a platform reselling tokens can't credibly offer, and it's why the choice between an agent and a plain workflow is a governance choice first.
- The signature is the product. However good the agent, what you're really building is the permission structure that lets it spend money safely — caps, approvals, exception routing, an auditable trail. Build that first. It's what turns a clever demo into something a CFO will sign.
The founders who sold the SaaS and rebuilt it as agents made an implicit forecast: in supply-chain spend, the durable business was never the software — it was the work the software only assisted. $75M says the market now agrees.
