How ‘agri-AI’ technology is changing farming

Fixed to a cattle chute at Agri-EPI’s South West Dairy Development Centre in Somerset, a small monitor is analysing cows’ legs and feet.

Using AI and thermal imaging technology, the ‘hoof monitor’ can spot lameness weeks before the farmer – saving£300 per cow typically incurred through treatment, reduced yield and shortened lifespan.

The early intervention is also thought to improve welfare and reduce carbon emissions per litre of milk produced.

It’s just one example of the agri-AI that’s hit the headlines in recent months – be it sensors detecting livestock disease, autonomous robots surveilling crops or monitors that count the wingbeats of bees.

But behind all the hype, to what extent does AI genuinely have the potential to transform agriculture? Are businesses embracing it? And how can retailers and suppliers get involved to speed up adoption and reap the benefits of more tech-savvy farmers?

Evolution of AI

Tech has been a fundamental tool on UK farms since the 1980s, when rudimentary sensors were used for climate control or to detect reproductive readiness in livestock.

These were followed by automated feeding systems, wearable collars to track individual animal metrics and, by the 2010s, remote monitoring software that tracked health and environmental conditions in real time.

Latest advances in AI – including its ability to collect, analyse and even make recommendations based on vast quantities of data – have unlocked a swathe of new potential applications and benefits. So much so that McKinsey estimates AI could eventually contribute $250bn to global agriculture via cost savings, improved yield and additional sales.

The applications in agriculture can be broadly divided into three core pillars: mitigating risk; driving efficiencies; and creating additional revenue streams.

The first area spans sophisticated monitoring tools such as the hoof monitor. These can detect disease in crops or livestock, or forecast climate events. The second area comprises tools that speed up production or reduce error, such as precision farming bots and autonomous tractors. Finally, we have the tools that likely do a bit of both, and indirectly create an additional revenue stream.

At Treefera, for example, its AI-enabled platform collates a vast amount of data from producers. That data is then ‘sold’ in exchange for carbon credits, explains Gabrielle Bourret-Sicotte, its ‘chief evangelist and head of customer success’.

“What theme ties all these use cases together?” asks James Watson, partner at consultants Argon & Co. “That AI is most effective when accounting for real-world variability such as unpredictable weather, changing soil conditions and shifts in grain moisture levels – all of which have long been the farmer’s enemy.

“AI in agriculture is about helping farmers make smarter, data-driven decisions – which, until now, hasn’t been possible.”

Though the precise benefits vary wildly depending on the technology, it’s thought AI has the potential to reduce annual agricultural operating costs by more than a fifth (22%), according to Arkinvest.

“AI provides farmers with several benefits, including the ability to reduce overall inventory levels, minimising excess and obsolete stock, which helps prevent margin erosion and issues related to expired inventory,” says Tom Gregorchik, vice president for industry strategy at Blue Yonder.

“By optimising resource use and improving yield quality, AI contributes to increased profitability and sustainability in agriculture. Additionally, as farmers increasingly integrate AI into their operations, they can improve distribution, reduce waste and ensure that consumer demand is met. This can lead to improved productivity, less waste and enhanced profitability.”

That’s the business case on individual farms. But the benefits of sourcing from AI-equipped farms spans the entire supply chain. There’s the obvious shared upside of higher yields, lower costs and better-quality food on supermarket shelves – and much wider benefits are up for grabs, too.

To start with, AI equips farms with a level of real-time data on animal welfare and environmental conditions that even the most robust assurance scheme can’t match.

“If you’re a retailer relying on a farm assurance audit to tell you these animals are being looked after, they’re healthy and welfare is good, you’re basing that on one maybe two inspections per year,” says Johnny Mackey, associate director, stakeholder engagement lead for MSD Animal Health.

Assessors might show up on what looks like “a model farm” with gleaming power-washed facilities and “cows bedded up to their bellies in fresh straw” and tick off their checklists, he points out. But the data might tell a different story – of livestock lacking enough water or fighting to feed.

“These devices are looking at the animals constantly,” he adds. “It tells you the truth and doesn’t rely on subjectivity.” That gives retailers and suppliers substantial evidence to verify and support on-pack welfare claims. It’s a crucial benefit at a time when 84% of consumers say they’d be put off by any hint of ‘welfare washing’, according to 2024 research by The European Consumer Organisation.

Plus, as major suppliers face growing regulatory pressure to track and report on sustainability metrics – on welfare as well as wider environmental impacts like carbon emissions and deforestation – they’ll need the granular data AI offers to meet their obligations.

“Suppliers absolutely need our services,” says Bourret-Sicotte. “Everything is a synergy because the farmer is ticked off as, say, EUDR-compliant and then suppliers like Nestlé, Nespresso or Unilever are safe to use them.”

A win-win?

If it’s all such a win-win, the industry should be falling over itself to get its hands on this innovation. Yet that’s not quite the case.

For some established technologies, like sensors on livestock, adoption is pretty widespread. MDA Health estimates about 50% of UK farms are using some type of monitoring, for example, though they vary in levels of sophistication.

But when it comes to the type of AI being splashed across headlines, there’s greater hesitation. Much of it isn’t commercially available yet, of course. Early trials are housed at research institutes, innovation hubs or demonstration farms, like the South West Dairy Development Centre.

Historically, there’s also been a tendency for AI to be developed by tech providers and academics,with little input from farms, says Chris Lyons, innovation lead for agriculture at Innovate UK.

“AI and machine learning is only as good as the data you feed into it”

Judith Batchelar, Food Matters International

While that’s starting to change, a far bigger hurdle is price tag. “Five years ago, [a farm] would have had to hire AI scientists, data scientists and engineers, build their own models and train it on their ground truth data. That would cost millions – definitely not accessible for any farm, except the really large ones,” points out Bourret-Sicotte.

Those costs are coming down somewhat. “Now companies like ours provide AI models as a service, so we have all the risk and we’ve hired all the really smart people,” she says.

More on AI:

• Ocado to slash headcount of R&D team as it eyes AI productivity gains

• Food waste apps say sharing is caring… but do they work?

• AI in food innovation will lead to stupid, boring products

• What shoppers really think about self-checkout machines revealed

Advances in Compute – roughly speaking, processing power – like faster processing cards from the likes of Nvidia “have made the cost of AI much lower”, Bourret-Sicotte adds.

But the reality is, early agricultural adopters of these technologies are likely to be shelling out quite a bit in some cases – no mean feat in an industry with notoriously low margins.

“Technology goes through economies of scale – ie it needs to get more products out there to reduce its unit price – but you’ve still got to be a risk taker if you’re going to go out there and be an early adopter,” says Lyons.

That puts the tech largely out of reach for the UK’s roughly 30,000 small farms. “Many farmers, particularly smallholders, struggle to secure the capital needed for AI investment, and available financial support is often insufficient to scale up its use on a commercial level,” says Nandini Chakravorti, associate director for digital engineering at the Manufacturing Technology Centre.

“Regulations such as data protection laws, drone usage and food safety compliance further complicate AI adoption. Navigating these legal and operational requirements can be time-consuming and costly.”

Watson points to another hurdle. “Skills and training gaps can also slow the adoption of AI in agriculture, as farmers need practical training to interpret AI-generated recommendations and make informed decisions based on them. If AI tools lack user-friendly interfaces, they risk being ignored or misunderstood.”

Take ethical chocolate brand Luker. It equipped farmers with AI to improve analysis of cocoa bean fermentation – a vital stage in chocolate production that involves the pulp in the cocoa bean broken down by yeast and bacteria, giving it that all-important chocolate scent and taste. Despite the benefits, there was some hesitation across its Colombian supply base, admits director of digital transformation Daniel Basto.

“The biggest challenge was integrating it with our people and existing processes,” he says. “When we first introduced the AI-powered fermentation tool, it wasn’t widely adopted — mainly because those using it hadn’t been involved in shaping or training with it from the start.

“Without that understanding, the technology felt unfamiliar and there was a lack of trust in its ability to integrate into a well-established buying process.”

This scepticism isn’t always misplaced, either. “There’s been some snake oil-type stuff going on, like in any new market,” says Casey Woodward, founder of AgriSound. Plus, there’s an attitude of “if it’s not broken, don’t fix it”. That’s less the case with younger farmers or more corporate groups, who can be more open to innovation, he says, but there’s a hesitation to spend money on what can feel like a risky new piece of kit, regardless of all the lofty promises.

“Many farmers struggle to secure the capital needed for AI investment”

Nandini Chakravorti, Manufacturing Technology Centre

The other challenge – at least, when it comes to unlocking the benefits in AI for the whole supply chain – is data readiness. So says Judith Batchelar, director of Food Matters International.

“The single biggest barrier is data standards,” she says. Take the monitoring of greenhouse gas emissions on a farm. Currently, the data collected and put into a carbon calculator lacks standardisation.

“AI and machine learning is only as good as the data you feed into it to learn on,” she says. “And if you don’t give it good, standardised data you can create all sorts of challenges – it can emphasise errors or inaccuracies.”

Get it right, though, and there’s huge potential. AI is now being used to analyse the nuances of geography, soil and microclimate to gauge the likely impact of biodiversity interventions at speed, for example. But again, there is emphasis that “you need to have good data to start with”.

Collaborative effort

The reality – for all of these reasons – is that AI likely won’t have the forecasted impact on agriculture unless retail and supply partners weigh in. That’s why, when AgriSound sought to get its bio-acoustic device implemented, it bypassed the farms and went straight to the retailer they supplied.

“The main value add for the corporates is they’ve then got that data they can use to feed into their sustainability reports,” says Woodward. “So, in most cases, either the retailer or the processor or integrator will purchase the technology or the licenses for it and deploy that into the supply chain.”

Marks & Spencer, for example, has implemented AgriSound across 10% of its supply base. “We’re a small team, so being able to go to someone who’s got that aggregation factor makes it more efficient for us,” Woodward adds. It also helps achieve “critical mass of interest” needed to scale the tech up.

“AI in agriculture is about helping farmers make data-driven decisions”

James Watson, Argon & Co

At Treefera, too, the provider works in tandem with farmer-producers and their buyers. “We provide data to both of them, they work in tandem and we provide the transparency,” says Bourret-Sicotte.

For its part, MDA Health has advocated for a ‘fully connected industry framework’. That would combine more targeted regulation and policies, and a joined-up strategy for agricultural adoption across supply chains and market premiums. The idea is to incentivise take-up on farms and reflect the fact that retailers and suppliers are also reaping the rewards of AI and other technologies.

“What can never happen is a supermarket or whomever getting access to the data without putting some skin in the game and using that data to impose maybe penalties on their farmers,” says Mackey. “That’s absolutely what we don’t want to happen.

“The ideal scenario would be a partnership between the retailer, the processor supplying that retailer and the farmer supplying that processor and the technology provider,” says Mackey.

“And if the processor/retailer is putting some investment in there, that’s buying them a right to that data, isn’t it? That would be the understanding. The farmer’s getting a better deal on devices proven to help their business be more efficient and therefore more profitable. And the supply chain wins because they’re getting data that’s going to help them meet targets and verify their claims.

“It’s genuinely a win-win situation,” he sums up. The benefits of which arguably won’t be accessible if farmers are left to their own devices, though.