Investors

Billions are pouring
into American dairy.
The farm still runs on a whiteboard.

Snowball builds FarmHand, an AI agent for the working dairy. It sees what the farm knows, reasons across it, and turns it into the day's work, with the right tasks sent to the right people. We're building the intelligence for an industry that's rebuilding itself.

In production Used every day by the crew of a working robotic dairy
2 partners First farm running now; a second, larger operation onboarding next
EN · ES Full interface and voice capture, both languages, live

The moment

US dairy is growing, consolidating, and re-tooling. New plants, near-record exports, bigger and better-capitalized farms. All of them are under the same pressure to produce more with less labor, and more willing than ever to pay for an edge.

$11B+
New US dairy processing capacity being built through 2027
$9.6B
2025 US dairy export value, up 15%, just shy of the record
68%
Of US milk comes from farms with 1,000+ cows, and rising
$3,130
Record replacement-cow price, April 2026. Every cow has to do more
  1. Dairy Herd Management, “The $11 Billion Redesign of the American Milk Map” (June 2026)
  2. USDEC, 2025 US dairy export results (February 2026)
  3. RaboResearch, 2026 World Dairy Map (consolidation and trade share)
  4. USDA via Progressive Dairy, replacement dairy cow prices (April 2026)

The gap

The industry modernized.
The farm office didn't.

A modern dairy has robotic milkers, sensors, and herd software, and it still runs its day out of the operator's head. A whiteboard, paper protocols, intuition, and a crew coordinated at shouting distance. Labor costs climb every year, and the appetite to optimize has never been higher.

Another dashboard won't close that gap. What a farm needs is something that sees everything it knows, reasons across it, decides what needs doing, and gets every task to the right person. That's FarmHand.

Why dairy first

The farm that never stops
is the farm that needs us most.

№ 001

The work is constant

Dairy runs around the clock, every day of the year. Milking, feeding, herd health, maintenance. There is no off-season to hide a labor gap in, and no farm where a missed task costs more.

№ 002

The hardware is installed

Robotic milkers, sensors, collars, and automated feeders are already on the farm, bought, bolted down, and underused. Dairies don't need new machines. They need software that makes the existing ones work together.

№ 003

The data compounds

A dairy makes hundreds of small decisions a day, and every proposal the agent makes is confirmed or corrected by the people who know the farm. That is labeled training data on how real operations run. Nobody else is collecting it, and it compounds every shift.

What exists today

Not a deck.
A working product.

FarmHand is an AI agent that runs a dairy's daily operating loop, and it is in production on a working dairy today. The crew clocks in on it, gets its work from it, and talks to it in English and Spanish. A second, larger operation is onboarding next.

  • 01 Workers and managers talk to it in plain language. An agent built on Anthropic's Claude turns what they say into work orders, and rules you can read decide who gets each one.
  • 02 It runs alongside the farm's herd records and robot systems (DairyComp, Lely), with direct reads rolling out so the agent can raise work of its own.
  • 03 Every action and every decision is recorded permanently, with the reasoning behind it. That history is the raw material our decision models learn from.
  • 04 Hard limits sit between the agent and the world. Approval tiers, spend caps, and a record of every decision it makes. Autonomy with a paper trail.

FarmHand, in the field now.

The long game

From running the crew
to running the farm.

Today

FarmHand runs the day's work. It captures what the farm knows, decides what needs doing, and sends each task to the right person. For now, human hands are the agent's actuators.

Next

It learns the operator. Every confirmation and correction is a trace of the farmer's judgment, and those traces train decision models that reason about the farm the way its best manager does.

Then

The actuators change. Robots, drones, and autonomous machines take tasks the same way the crew does today, and the whole farm keeps moving with the farmer directing, not driving.

Every farm modernizes at its own pace. We meet each one where it is.

Why now, why us

The technology just arrived

Farms always needed software you could just talk to. Until frontier language models, it didn't exist. Now it does.

Nobody sees the whole farm

Farm data sits stranded across herd software, robot systems, and paper. The agent that reasons across all of it becomes the smartest thing on the farm, and the hardest to replace.

Trust is the product

Agents that touch real operations need audit trails, approval tiers, and hard limits. We built those before anything else. It's the difference between a demo and a business.

“…will evolve toward multi-agent ecosystems coordinating full-farm operations… the goal is not to replace farmers' judgment but to extend it.”

— JDS Communications, a journal of the American Dairy Science Association, on agentic AI in dairy (2026)

Get in touch

Let's talk.

If you're investing in agriculture, applied AI, or the labor economy, we'd like to show you what's running. The full evidence pack is available on request.