AI that knows your farm.
Built to actually use.
Three classes of AI, built into the platform: a conversational advisor that knows your operation, an agentic orchestrator that can do the work for you, and computer-vision pipelines that grade quality, spot disease, and recognise individual animals. All grounded in your own records — with the audit log and approval gates a working farm requires.
Advisor, Orchestrator, Vision.
Each one does one thing well, and they share the same ledger so the advisor's answers, the orchestrator's actions and the vision pipeline's results all live in the same record set.
Farm Advisor
Conversational, knows your paddocks, sprays, mobs, observations and compliance status. Ask anything — get answers grounded in your records.
Master Orchestrator
50+ tools across 18 modules. Books transport, logs sprays, files compliance records, posts ledger entries — with per-action approval gates and a tamper-evident audit log.
Vision Intelligence
Point your phone, get a result. Grain & cotton grades, pest and disease ID with severity, body condition scores, ear-tag reads, animal recognition — all returned as structured data.
A 24/7 assistant that actually knows your farm.
Not a generic chatbot bolted onto a help centre. The Advisor is grounded in two layers: the platform's own knowledge of how everything works, plus your live farm data — paddocks, crops, recent sprays, observations, harvests, compliance status.
Ask it how to do anything in the platform
“How do I record a spray with delta-T?” “Where do I see my myBMP progress?” “Show me last quarter's stock movements on the south block.” The Advisor answers with the same precision a trained team member would — because it has the platform's full operating guide cached as context.
Ask it anything about your operation
“What was the spray on Paddock 4 last week?” “Which mob is overdue for drench?” “What's our exposure if cattle drop 20c?” Answers come from your own ledger with the source records linked — not a guess from a model that's never seen your farm.
Always-on, paddock-ready
Works on the tablet at the chute, on the phone in the ute, at the kitchen table. Available to every worker on your tenant, with the same answers, all the time.
Cost-efficient by design
Anthropic prompt caching means the platform's knowledge base only costs tokens once. Your follow-up questions only pay for the new bits — so the Advisor stays usable for routine questions, not reserved for "important" ones.
An agent that does the work — with the brakes on.
The Orchestrator runs on a curated catalogue of 50+ tools across 18 operational modules — livestock, dairy, breeding, spray, compliance, marketplace, accounting, water, weather, transport and more. It can take action on your behalf, with the controls a working farm and its auditors require.
It actually does things
“Book transport for the south mob to the saleyards Friday.” “Log this morning's spray on Paddocks 4–7.” “Mark today's tasks complete and roll the unfinished ones to tomorrow.” “File the LPA record for the treatment we just logged.” The Orchestrator drafts, you approve, the action happens.
Every tool has a gate
Each tool is set per tenant to one of three modes — auto-run (safe, idempotent reads or low-stakes writes), needs approval (anything that touches money, livestock or compliance — surfaces a confirmation card you sign off), or blocked (disabled for this tenant entirely). You choose where the line sits.
Tamper-evident audit log
Every tool invocation is logged: who triggered it, when, what arguments were passed, what came back. The log feeds straight into your audit pack — “the agent did this, the human approved it, here's the record” — so an AI-assisted operation stays defensible to auditors.
Kill switch in Settings
One toggle disables the entire agent for your tenant — instantly, no rollback, no support ticket. If you ever want to pause it for a quarter or shut it down completely, that's a single click.
Point your phone. Get a result.
A computer-vision pipeline (Vertex AI Gemini 3 Flash under the hood) that turns photos into structured agronomic data. Take the shot in the paddock; the dashboard updates the moment the result lands.
Crop quality grading
Grain: GTA grade (Prime Hard → Rejected), defects detected, moisture estimate.
Cotton lint: USDA colour grade, micronaire, staple length, fibre strength, leaf grade, plus a contamination breakdown (bark, plastic, twine, oil).
Crop health & pest ID
Verticillium wilt, fusarium, black root rot, helicoverpa, aphid, mite, silverleaf whitefly — each returned with a severity (critical / high / moderate / low) and an array of co-occurring issues. A paddock walk becomes a structured report.
Livestock at a glance
Breed identification (primary + secondary for crossbreds), Body Condition Score (1.0–5.0), weight estimate in kg, ear-tag OCR (reads NLIS / management numbers from the photo), and health flags — lameness, pink eye, bloat, fly strike.
Results flow into your records in real time — dashboards subscribe to the vision pipeline and update the moment a frame is graded.
Recognise individual animals — without an ear tag.
Each animal's distinctive markings produce a stable visual fingerprint, stored as a vector. Photograph the same animal weeks later — the system returns the same identity match, even if the tag is missing, the RFID's failed, or the animal's never been in the chute.
RFID-tagless herd management
Build and maintain a herd register from photos alone. Useful for studs, organic and free-range operations where tags are impractical, or as a redundancy layer for conventional tagging.
Lost-animal recovery
A neighbour finds a stray and snaps a photo. Run it against your herd's fingerprints; if it's yours, you'll know in seconds — no chute work, no scanning, no guess.
Drone-image mob counts
Fly the paddock, feed the imagery to the platform, get an automated count with individual matches against the register. Far less labour than a manual head count and a lot more accurate than a clicker.
Messy input, sorted automatically.
A Claude-powered inference layer that takes whatever a worker types into a phone in the field and routes it to the right event, the right module and the right compliance record — so you don't lose data to a rigid taxonomy.
Free-text to structured event
A worker types “top-dressed paddock 4” on their phone. The classifier files it as a fertiliser event against Paddock 4, in the crop module, with the correct event type — no dropdowns, no taxonomy training.
Auto-categorised expenses
Invoice descriptions get mapped to the right ATO / MTD expense category automatically. Reconciliation goes from a manual chase to a review queue — much smaller surface to actually correct.
SFI / programme validation
Activity descriptions are checked against Sustainable Farming Incentive criteria (and equivalent programmes elsewhere) so you can see, before you submit, whether what you've logged actually qualifies.
Compliance enrichment on sync
When records sync up from the field, the classifier adds the compliance metadata they're missing — withholding period flags, audit codes, programme tags — so the records arrive in a form the audit pack can use directly.
Goodbye blank page.
When you open the Hiring module to post a seasonal job, the platform drafts the role description, the requirements and a day-in-the-life — based on your farm's actual operation. Edit it, post it, move on. Same approach is rolling out to other places we ask you to write from scratch.
Forward-looking what-ifs against your real P&L.
Model a decision before you make it — add a paddock, drop a commodity price, change the rotation, buy 200 head. The Scenario Planner runs the simulation against your live records, side-by-side with the actual financials, without touching the working data.
Built on your actual numbers
Every scenario starts from your real P&L, balance sheet and operating records. The only thing the simulation changes is what you ask it to change.
Compare side-by-side
Run a scenario, see the impact on cash, debt service coverage, biological asset position and gross margin against the baseline. Save it. Run another.
Sandbox by design
Scenarios never affect your live ledger or compliance records. Try as many variations as you want — your real books stay untouched.
Share with the bank, the buyer, the accountant
When the bank asks “what if”, you don't reach for a spreadsheet — you send them the scenario. It's the same shape as the records they're already seeing.
Built so you can actually use it in production.
AI on a working farm has to be auditable, controllable and reversible — or it can't be used at all. The platform's safety architecture is what makes the rest of it deployable.
Per-tenant tool gates
Every action the agent can take is auto-run, needs-approval, or blocked — set by you, per tool. The defaults are conservative; you turn things up as you build trust.
Approval cards for consequence
Anything that touches money, livestock or a compliance record surfaces a confirmation card with the proposed action laid out. You approve in one tap — or reject, and the agent moves on.
Full tamper-evident audit log
Every prompt, every tool call, every approval, every outcome. The log is signed into the same Merkle root as your operational records, so AI actions are as auditable as anything else on the platform.
Permissions, not exfiltration
Inference runs against your tenant's data inside our boundary. Your records don't leave to power somebody else's model, and the AI doesn't see other tenants' data.
Want to see it run on your farm?
Early access is opening in waves. Join the waitlist and we'll show you the Advisor, the Orchestrator and the vision pipeline working against real records.
Early access · 30-day free trial · No credit card required