Our AI does not need to learn your horse first

Ask how a behaviour alerting system works and you usually get the same answer: it watches the horse for a while, works out what is normal for him, and warns you when he departs from it.

It sounds sensible. It has a hole in it, and the hole is exactly where you need the system most.

The problem with learning a baseline

A system that has to learn a horse's normal needs healthy, uneventful days to learn from. Three consequences follow, and none of them are academic.

• The horse is unprotected during the learning period. The nights you paid for are the ones the system spends taking notes.

• It fails on a horse that is already in trouble. A clinic patient arrives sick. There is no healthy baseline to learn, so there is nothing to compare against.

• It has to be installed well ahead of the season, which is exactly when nobody has time to install anything.

TEO is pre-trained

NOVOSTABLE runs on TEO, an AI trained on more than 25 billion records and over 1,000 real cases: colics, foalings, foals, horses cast against a wall. It has already seen what these things look like, on a very large number of horses.

So it does not arrive at your yard as a blank page waiting to be taught. It arrives knowing what a horse in trouble looks like.

Live on any horse in about 30 minutes

Thirty minutes of presence in the box is what it takes for TEO to be working on a horse. Not thirty minutes to learn that horse, there is no per-horse learning period at all. Thirty minutes to be operational.

In practice that removes the planning problem. A mare who arrives this afternoon is covered tonight. A patient admitted at six is covered by half past six.

It works on a horse that is already abnormal

This is the case that separates the two approaches, and it is decisive for clinics. A horse under a drip, a post-operative case, a horse admitted in the middle of a colic: none of them will ever produce the healthy baseline that a learn-the-normal system requires.

TEO does not need one. It reads the behaviour against everything it has already been trained on, not against a history of that individual horse.

What it measures, and what it does not

TEO measures behaviour, every minute, around the clock. Posture, movement, how long the horse is down, how he gets up, how he moves in the box.

It does not measure health. It does not score welfare. It does not diagnose. Those belong to the vet, and a system that claims otherwise is selling something it cannot deliver. What the AI does is narrower and more useful: it recognises abnormal behaviour, and it tells you.

Why it calls you

A push notification at three in the morning is a notification nobody reads. A phone call wakes people up. An alert is only worth what it triggers, and at night only a ringing phone triggers anything.

Nobody sits in front of a screen. There is no rota to fill, no shift to staff, no wall of live feeds to watch.

The camera is a sensor, not the product

The camera in the box is an input, in the same way a thermometer is an input, and it is very often the camera you already have. What you are putting in place is not video surveillance. If the value were in the camera, anyone could sell it to you.

What this looks like across a season

• A mare foaling: the behaviour shifts before the waters break, and the call goes out with enough time to walk down to the box.

• A horse starting a colic at two in the morning: pawing, up and down, turning to the flank. The phone rings.

• A horse cast against the wall and unable to get up.

• A clinic patient on a drip overnight, with nobody in the building.

See what it would look like on your yard

The useful question is not what an AI can do in general. It is what it would see in your boxes, with your cameras, on your horses, starting tonight. Book a demo and we will go through exactly that.