AsureQuality Case Study

Food quality assessor's AI prototype promises fatter margins for meat growers

AI is famous for augmenting human-centred processes, but Government-owned AsureQuality has bigger plans that could deliver better returns to beef and sheep farmers. A successful proof-of-concept shows just how.
Big idea: AI in the abattoir could detect disease and reduce wasted animal flesh, fattening farmers’ wallets.
Big Idea
AI in the abattoir could detect disease and reduce wasted animal flesh, fattening farmers’ wallets.
Approach: Develop and evaluate an AI concept combining video and a Computer Vision AI stack that has good eyes for disease detection.
Approach
Develop and evaluate an AI concept combining video and a Computer Vision AI stack that has good eyes for disease detection.
Brains behind the project:  Matt Scott, AsureQuality’s Inspection Innovation Manager, worked with consultants at VISIR.
Brains behind the project
Matt Scott, AsureQuality’s Inspection Innovation Manager, worked with consultants at VISIR.
3 Months
From concept design to proof-of-concept completion.

Timeline

3 months from concept design to proof-of-concept completion.

Achievements

Initially developed for cattle, the model now also applies to sheep. The model is being described as the first of its kind in the world.

Next Steps

Incorporate virtual reality to guide and review abattoir worker excisions. Train the model to detect other diseases and contamination. AsureQuality could license the offering for deployment in other countries.
Achievements
Initially developed for cattle, the model now also applies to sheep. It is funded for further development. Food safety experts from around the world love what they see. The model is being described as the first of its kind in the world.
Next steps
Incorporate virtual reality to guide and review abattoir worker excisions. Train the model to detect other diseases and contamination. AsureQuality could license the offering for deployment in other countries.

The world expects a lot from Kiwi food producers

The world can’t get enough of New Zealand food. Dairy export revenue is nudging past $27 billion, meat and wool exports are around $12.3 billion, while horticulture is expected to surpass $8.5 billion in export sales this year. New Zealand is great at growing food and trusted for its methods and quality. AsureQuality is a big part of this story, for 150 years monitoring the quality and safety of food exports, helping to grow the country’s global reputation in food production.
A big chunk of the organisation’s work involves meat and abattoirs. A peek inside shows lines of inspectors in white coats dotted along the killing chain, checking animal carcasses for disease or contamination. It’s a critical part of the process and the difference between successful export trade partnerships and cancelled orders.
Matt Scott, the automated inspection lead at AsureQuality, is exploring how AI can assist meat inspectors and fatten returns to farmers. Rather than simply automate elements of meat inspection, Mr Scott is taking a broader perspective that will also benefit farmers and the organisation itself.
“It feels like the world is looking to New Zealand to develop the technology in this space, because we’re trusted.” - Matt Scott

Camera, Action, AI

AsureQuality worked with local AI consultancy VISIR to design and deploy an AI concept in three abattoirs to see what was possible. The project aimed to develop a model that could detect peritonitis (inflammation in the lining of the abdominal cavity) and pleurisy – another form of inflammation that can lead to adhesions between the lungs and chest wall – in slaughtered cattle (sheep would come later). The presence of either condition, though rare in New Zealand, can lead to carcass downgrades or even condemnations.
Video cameras focus on the chest cavities of animal carcasses moving on overhead rails through the abattoir. The footage is stored on a local computer. At the end of each shift, video is exported to AsureQuality’s network infrastructure and subsequently transferred to a private storage bucket within the organisation’s AWS cloud environment. The entire pool of video connects to a LLM (Large Language Model) stack, which the project team trained to recognise disease states. Algorithms also anonymise footage, screening out abattoir workers (even blurring name tags stitched into overalls) and surroundings to focus purely on animal carcasses.
Video cameras focus on the chest cavities of animal carcasses moving on overhead rails through the abattoir. The footage is stored on a local computer.
At the end of each shift, video is exported to AsureQuality’s network infrastructure and subsequently transferred to a private storage bucket within the organisation’s AWS cloud environment. The entire pool of video connects to a LLM (Large Language Model) stack, which the project team trained to recognise disease states. Algorithms also anonymise footage, screening out abattoir workers (even blurring name tags stitched into overalls) and surroundings to focus purely on animal carcasses.
Given the scarcity of disease in New Zealand farm animals, the model has had to process a massive volume of video (1.7 million 10-second segments at the time of writing) to get up to speed. Synthetic data – images that have been doctored to display disease where there is none  - was used to further enhance the model’s detection capabilities. Meat inspectors play a role, too, logging in to a private portal to review and annotate footage that shows telltale signs of disease.

Trimming Waste

Animal carcasses with disease are diverted off the killing chain, sectioned, and further examined and reclassified. Carcasses meeting specific standards have localised infections removed. Workers in this part of the killing chain often follow the skin surgeon’s rule: "if in doubt, cut it out."
While a generously broad knife strokes minimise the risk of tainted meat ending up where it shouldn’t, too much good meat ends up on the cutting floor. It’s a big problem – and an awful waste – when you consider that some plants process over one million lambs a season.  With a bit more training, AI will help fix that, Mr Scott said.
Animal carcasses with disease are diverted off the killing chain, sectioned, and further examined and reclassified. Carcasses meeting specific standards have localised infections removed. Workers in this part of the killing chain often follow the skin surgeon’s rule: "if in doubt, cut it out."  While generously broad knife strokes minimise the risk of tainted meat ending up where it shouldn’t, too much good meat ends up on the cutting floor. It’s a big problem – and an awful waste – when you consider that some plants process over one million lambs a season. With a bit more training, AI will fix that, Mr Scott said.
“Sometimes way more is cut off than is necessary, so AI could provide a trimming guide that shows exactly which areas need to be removed, how deep and wide slaughtermen need to cut, and that – actually – you don’t need to remove the entire rib cage,” he said. “A 1kg rack of lamb retails for over $50, so improved cutting accuracy will make a real difference to farmer yields.”
Mr Scott envisages a future where abattoir workers will use augmented reality headsets to  overlay cutting lines on the infected parts of animals. The tech combo could also assess the cutting handiwork of workers, scoring their accuracy, offering a powerful training aid.
But it’s not processors and farmers who are rubbing their hands together at the prospect of AI in the abattoir. “There’s a very low incidence of disease in New Zealand and some people struggle to see the value in what AsureQuality does,” Mr Scott said. “It’s easy to view us as a bit like insurance – a grudge purchase, a kind of red tape on exports,” Mr Scott said.
“AI gives us the opportunity to reframe what we do and add much more value.” - Matt Scott
“AI gives us the opportunity to reframe what we do and add much more value.” - Matt Scott

Emerging Possibilities

Faecal contamination is more likely than disease to endanger export orders. Tiny traces can turn around a fully laden container and put the country on a blacklist for months. Inspectors rarely miss the signs, but AI could add another set off eyes to further reduce risks. Incorporating alerts will help inspectors catch any contaminants early and pinpoint necessary changes on the killing chain. The organisation will also train the model to detect upwards of 10 other disease states.

Global Opportunity

Mr Scott recently presented the proof of concept to government  advisors to pave the wave for the next phase of development and put AsureQuality at the vanguard of AI in food safety and monitoring. He said several food processors in other countries were in the early stages of AI development but lagged behind AsureQuality’s current model – a view confirmed after it was presented at the annual meeting of the International Heads of Food Agencies Forum (IHFAF), held in Santiago earlier this year.
“I believe the feedback from that presentation made it clear New Zealand is leading the world in AI’s application to food safety,” Mr Scott said. “What we're doing transcends markets and countries – we can offer the New Zealand standard of food safety to other countries. The model will also serve as an instrument to support our reputation as the leader in food safety – now we’ve got hard data to say, ‘You really can trust us’”. 
AI is automating processes and making businesses smarter – and it’s happening faster than you think. Don’t sit on the sidelines. Good working prototypes can take as a little as a few months to get off the ground and don’t cost a fortune.
Contact VISIR now to see what we can do for your business!  
Contact Us
Contact Us
Company
Our ProcessOur TeamCase Studies
Resources
Visir Wins "Best Places To Work" Award
See all blog posts >
Copyright © 2025 VISIR GROUP LIMITED