TL;DR
Marks and Spencer has rolled camera-based welfare monitoring across every farm in its milk pool, 46 in total, following trials at a smaller group. Video of housed cattle is scored by a model built with veterinary input, and the resulting data has already prompted changes to bedding, feeding times and stall design.
What the system measures
Each animal gets a running health profile rather than a snapshot at inspection. The analytics track how long a cow lies down, how long she spends feeding, how she moves and how she interacts with the herd, combining those into a comfort score for vets to work from. Separately the cameras estimate body condition, the standard measure of nutritional state, and follow how it shifts across weeks.
That second measure is the one with clinical value. Terry Canning, chief executive of supplier Cattle Eye, notes that body condition moves before metabolic and fertility problems become visible, and that tracking it herd-wide at frequency was previously confined to research settings. On testing, the algorithm reportedly matched the judgement of a trained welfare assessor.
What changed on the farms
The trial results are unusually concrete for an AI deployment. Farmers who saw lying times fall adjusted individual animals’ bedding to suit their size. Feeding was rescheduled to cut the hours cows spent standing. Some stalls were physically rebuilt. None of that is a model output; it is a person acting on a measurement they did not previously have.
Peter Kennedy, who heads agriculture at M&S Food, ties the rollout to the retailer’s farming resilience programme. Map of Ag veterinary consultant Jenny Allan points out that the welfare focus predates the technology by well over a decade — the cameras extend an existing regime rather than establishing one.
Looking forward
Most retail AI announcements this year have concerned the customer-facing end: Morrisons trialling scan-free trolleys in Preston, chatbots on product pages, assistants making claims the advertising regulator has now written rules about. This one sits in the supply chain, where the output is an audit trail rather than a conversation.
The commercial logic is straightforward. Welfare standards are a differentiator British supermarkets already compete on, and continuous measurement produces evidence that periodic inspection cannot. The open question is what happens when the data shows a supplying farm falling short, and whether a score generated by a camera becomes grounds for delisting.