Know what your pasture will do next — we're looking for farmers to help build it
Paddock-level growth predictions, tuned to your farm and your own observations.
On a 2,000-hectare property, the pasture data you can trust is the data you rarely have time to collect. Walking covers paddock by paddock — plate meter out, week after week — is a serious slice of someone's day, and it's the first job to slide when everything else is going on. So the feed decisions that hang off that data end up running on a look out the ute window and whatever you're carrying in your head.
That head model is good — often better than any satellite. But it lives in one person, it's hard to hand to staff or a manager, and it gets stretched thinnest in the very seasons (a dry, a wet spring) where a wrong call costs the most: your cheapest feed gone to seed on one side, stock losing condition and the chequebook out on the other.
Atamai is developing paddock-level pasture availability and forward growth prediction with one aim in mind: get the model accurate enough on your farm that the measuring folds into what you'd do anyway when you shift stock — not a separate monitoring round on top of the work. It sits in one place alongside your compliance and stock records, rather than as yet another standalone app to log into and reconcile by hand. And the part we care most about getting right is this: the prediction should fit your farm specifically, and treat your own eyes as the source of truth.
Why "fits your farm" is the whole point
Satellites are good at telling you what a paddock looks like today. Weather data is good at telling you what's likely to happen next. Neither one knows that your river flats come away faster after rain than your neighbour's, or that a particular block always lags in spring. Your farm is not the average farm, and a one-size-fits-all model leaves accuracy sitting in the paddock.
So the model calibrates to each farm. Your plate-meter readings, a quick phone photo of a sward, soil sensors if you run them, and your own notes all feed back in — and the prediction gets more accurate on your country over time. The technical problem of estimating pasture from an image is largely solved in the research world. The problem worth solving — and the one our R&D programme is built around — is making a model that keeps adapting to a real, changing farm, season after season, across a whole fleet of farms at once.
We'd rather agree with you than override you
A satellite image is a proposal. A growth prediction is a proposal. You accept it, adjust it, or bin it against what you can actually see. We're building the product to respect that, not fight it:
- Predictions come with honest uncertainty. We'll show you a range, not a single number dressed up as fact. Early on — before the model has much history on your farm — that range will be wider. That's the system working, not failing.
- When our estimate matches what you see, you'll see that too. The whole thing is built around the loop between your observations and ours closing over time. That's how trust gets earned: not by us being right, but by us repeatedly agreeing with what you already knew.
- When we disagree with you, you win. A mismatch isn't a farmer error to confirm away — it's information that retrains the model for your farm. The system suggests; you decide. Always.
Your data stays yours
This matters, so we'll be plain about it. Participating in the research programme means sharing some of your pasture observations with us so the model can learn. It does not mean signing your farm data over to anyone.
You own your data. Participation is opt-in and you can withdraw. We don't sell farm data — our business is built on giving you the tool and giving your chosen partners verified compliance access with your consent, not on trading what's on your farm. And if you ever leave the platform, your calibration history leaves with you.
The research consent spells all of this out in full — exactly what we collect, what it's used for, how long we keep it, and how you pull out. Please read it before you sign up.
What being a research participant looks like
We're recruiting a small group of sheep and beef farmers to help calibrate and pressure-test the model through a full season. This first season is the work of getting it accurate on your country — so in practice it means capturing the odd plate-meter reading or pasture photo as you shift stock, and telling us when the prediction is off. The aim is that the measuring keeps shrinking as the model learns your farm, not that you take on a new monitoring chore. In return you get early access to the pasture tools, a model that's actively tuning itself to your country, and a direct line into how the product gets built.
If that sounds like your kind of thing, we'd love to have you in.