Home/AI assistant/Paddock Notes
AI assistant · Paddock Notes

Talk. It sorts out the record. Being finished

Speak the way you actually talk in the yards — “drenched 4471 and her calf, marked 3921 a poor doer, check the fence in the north paddock” — and get three clean, separate records to confirm with a tap.

The reason paddock work doesn’t get recorded is that stopping to type it is a pain when your hands are full and the race is backing up. So you remember it. Or you don’t. Paddock Notes lets you just say it — one messy, off-the-cuff sentence — and turns it into properly separated records: a treatment here, an observation there, a job for the fence.

It’s built to beat a rigid voice-command grammar, not match it. You don’t have to learn the magic words or say things in order. Multiple animals, multiple actions, and an aside about the fence, all in one breath — it untangles them, attributes each to the right beast, and hands you clean records to check. You confirm; nothing is saved behind your back.

How it works

From one spoken sentence to clean records.

01

Speak it

Talk into the field app in the yard. A rough transcript is made on the device, so it works with no signal.

02

It untangles it

On sync, the assistant separates the animals and actions into distinct records — treatments, observations, even a job.

03

You confirm

Each record is shown for a one-tap confirm or a quick edit. Nothing is written until you say yes.

Handles real, messy speech

“Drenched cow 4471 and her calf, marked 3921 as a poor doer, and check the fence in the north paddock” is three different things about three different subjects. Paddock Notes pulls them apart: a drench treatment on 4471 and its calf, an observation on 3921, and a job on the north paddock. You didn’t have to structure it — you just said it.

Works with no reception

The rough transcript is captured on the device the moment you speak, so a flat spot with no signal is no obstacle. The tidy-up into structured records happens when you’re back in range and sync — and because everything else about the app is offline-first, the resulting records queue and reconcile just like anything you’d tapped in by hand.

It asks rather than invents

If a tag is unclear, a dose is missing, or “her calf” could mean two animals, Paddock Notes flags it for you to resolve instead of guessing. The stakes are your real records, so an ambiguous note becomes a question, never a made-up animal ID or a fabricated dose. You are always the one who confirms what gets written.

What’s being finished

Paddock Notes is in final build. The offline transcript, the multi-entity extraction, and the confirm-before-save flow are described here as they’ll work when complete — the point at which we onboard your team. We’re hardening the disambiguation on the trickiest notes, and adversarially testing the guardrail so it declines or asks rather than inventing anything on an incomplete note.

The hard rule

Nothing is ever written from a note without your explicit tap to confirm. The draft shows every field it would record, and an ID or a dose it could not resolve from your own records is dropped rather than guessed.

Reference

The detail.

InputFree-form spoken notes in the field app
OfflineOn-device transcript; structured extraction on sync
HandlesMultiple animals and actions in one note; asides become jobs
SafetyConfirm-before-save; asks when ambiguous; never invents IDs or doses
CreatesTreatments, observations, and jobs — each to the right subject
StatusLive in the web app and the field app

See it on your own place.

Start free on up to 200 head — no card, no lock-in. Running a bigger show? Book a demo and we’ll set it up around how you work.

/* RESPONSIVE PHOTOS. A phone was downloading the full 1920px desktop original to paint a 390px band — 364 KB for the hero alone, on an audience defined by rural mobile data. The small file is the DEFAULT and the bigger ones are opt-in by width, so anything that matches no query still gets the light one. Variants generated with sharp at q78: 760w and 1280w. Regenerate them if the source photos change (node -e with sharp; see the plan entry for 2026-09-20). */ @media (min-width:761px){ .ph-herd { --photo:url(/img/cattle-herd-1280.webp); } .ph-grass { --photo:url(/img/cattle-grassland-1280.webp); } .ph-dairy { --photo:url(/img/cattle-dairy-1280.webp); } } @media (min-width:1281px){ .ph-herd { --photo:url(/img/cattle-herd.webp); } .ph-grass { --photo:url(/img/cattle-grassland.webp); } .ph-dairy { --photo:url(/img/cattle-dairy.webp); } }