High Growth AI Accelerator - Agriculture

Hartpury University and Hartpury College has 4,600 students from more than 60 countries studying PhDs, postgraduate and undergraduate degrees, and diplomas in specialist subjects including agriculture, animal, equine, sport, veterinary nursing, business management, and A-levels. As a key player in the agri-tech sector, Hartpury is home to a Digital Innovation Farm that provides real-world testbeds for smart farming solutions, precision livestock technologies, and data-driven decision-making. Through collaboration with industry, Hartpury supports the development and adoption of cutting-edge technologies and automation to improve productivity, sustainability, and resilience in the agricultural sector. The institution is committed to ensuring students are well prepared to take advantage of technological developments in their future careers.

Biodiversity monitoring

How might we enhance bioacoustics detection of different wildlife species for enhanced biodiversity monitoring nationally?

Biodiversity across the UK landscape has declined sharply in recent decades. Widespread, real-time monitoring of wildlife, particularly birds, bats, and insects, is needed to understand ecosystem health and inform land management. Tracking the sounds of different wildlife, or bioacoustics, have been proven as a reliable method for surveying habitats, and acoustics datasets are becoming more widely available from national survey work like Wildlife Trusts. Monitoring wildlife on a large scale with this technology over time could be hugely beneficial for enhancing the UK wildlife population and reversing biodiversity decline. However, further calibration, validation and deployment of bioacoustics sensors is required to enable this.

Bioacoustics sensing offers a non-invasive, scalable way to monitor biodiversity, but classification models need improvement to distinguish species accurately across varied landscapes. As wildlife sound data is becoming more readily available via databanks, Hartpury Farm is seeking AI/ML tools to process and interpret these recordings at scale

Ideal solutions will:

• Detect and classify a wide range of species from environmental audio.
• Leverage national datasets to improve model accuracy and generalisability.
• Prove that actionable data can be extracted for landowners, ecologists, and policymakers.

Methane monitoring for livestock

How might we improve analytics capabilities utilising methane emissions from individual ruminant livestock, such as cattle?

Methane from ruminant livestock, a by-product of food consumed, is a major contributor to agricultural greenhouse gas emissions and, contrary to popular belief, is the more significant source when compared to animal flatulence. While reducing these emissions is key to sustainable livestock farming, the measurement of methane output at the individual animal level is complex and rarely done on-farm.

This challenge aims to develop an AI-powered software solution that can detect and analyse methane emission patterns using data collected from mobile gas analysers. These analysers measure exhaled methane during routine farm activities like weighing or milking. The opportunity lies in building scalable analytics tools that operate alongside these hardware systems to inform nutrition, breeding, and emissions reduction strategies.

Proposed solutions should:

• Detect and classify methane belch events from sensor data and integrate with existing farm workflows and data systems.
• Generate insights that support interventions such as dietary changes or selective breeding.
• Be designed for commercial scalability, with a focus on modular, cloud-based, or edge-compatible deployment.

Peacock Technology is a global pioneer in the application of advanced machine vision and automation technologies to solve challenging tasks in real-world environments. It specialises in Intelligent Dairy Solutions, leveraging robotics and artificial intelligence to achieve efficient and sustainable dairy production. Committed to innovation, the company develops AI-powered solutions that enhance dairy cow productivity, improve animal welfare and facilitate more eco-friendly livestock management practices.

3D modelling and behavioural tracking in livestock barn

How might we create a data-rich digital twin of a livestock barn to enable predictive simulation as a change management tool to encourage improvements in farm practices?

Historically, farms have relied heavily on manual observation to monitor things like animal health and welfare, resulting in limited data to identify opportunities to improve operational efficiency. On the contrary, Precision Agriculture is a modern farming method which utilises sensors and data analysis to enhance efficiency, productivity, and sustainability. By enabling digital simulations of farm environments, it allows farm managers to monitor smaller areas or even individual plants or animals based on their specific needs.

Peacock Technology seeks to develop a high-fidelity digital twin of a farm facility, specifically a 120m x 50m cow barn, integrating existing sensor infrastructure, including a network of over 100 cameras. The goal is to go beyond basic surveillance and toward a virtualised environment, potentially using tools such as CAD models or real-time 3D game engines (e.g. Unity or Unreal Engine) to simulate and analyse animal behaviour. By integrating spatial modelling with AI vision analytics, the system could enable real-time monitoring of cow behaviour (movement, feeding, resting) and alerting, health prediction, and scenario testing for barn design and changes in farm practices for better animal welfare and productivity.

Peacock Technology wants to establish a framework for precision livestock farming that is scalable across other farm types or species. Therefore, this challenge seeks AI/ML solutions that can:

• Enable real-time behavioural tracking, and alerting, through advanced image recognition across a multi-camera setup.
• Integrate camera feeds into a unified 3D environment using game engines or simulation tools.
• Generate predictive analytics and simulate barn layouts or interventions to assess their impact on animal health, productivity, and welfare.

Visual livestock identification at scale

How might we develop a low-cost, camera-based system to identify thousands of animals on-farm using visual markers visible to standard IP cameras?

Traditional animal identification methods, such as RFID tags or manual records, often pose challenges in terms of cost, maintenance, and scalability, especially in large-scale farming environments. As precision agriculture expands to include welfare auditing and biosecurity, a more accurate and scalable livestock identification system could transform the way in which farms are managed.

Peacock Technology is interested in exploring a camera-based approach to livestock identification that uses low-cost, visually identifiable collar system via standard fixed-position IP cameras from up to 10 metres away. These collars might utilise high-contrast visual markers (e.g., ArUco-style glyphs, custom colour codes, or numeric IDs) that are easily readable by computer vision systems under varied lighting and weather conditions. The goal is to support farms managing populations of up to 15,000 animals, enabling individual identification for behaviour monitoring, welfare auditing, and automated access control, without the need for expensive or invasive sensors.

The key challenge is achieving high reliability and precision in real-world conditions, while minimising compute requirements to run at the edge or via efficient cloud pipelines and keeping material requirements affordable for mass adoption and real-world use. Therefore, this challenge seeks AI/ML solutions that can:

• Accurately detect and decode visual markers on animals from IP camera feeds across a variety of angles and environmental conditions. 
• Link visual IDs to broader data systems (e.g., feeding, health records, behaviour tracking).
• Offer extensibility across species or farm types, and adaptability to new marker designs or configurations.

Nestlé, one of the world’s largest food and beverage companies, is deeply committed to sustainable sourcing, supply chain resilience, and improving farmer livelihoods. With cocoa at the heart of many of its iconic products, Nestlé invests in long-term programmes across West Africa—where over 90% of Europe’s cocoa is produced—to enhance crop health, farmer income, and environmental sustainability. As part of its broader commitment to regenerative agriculture, Nestlé is also making major investments in agroforestry within the region’s cocoa sector. These initiatives aim to reduce carbon emissions, strengthen farm resilience, and promote long-term sustainability across global supply chains.

Nestlé’s R&D network has become a powerhouse of cocoa innovation: at Zambakro, its 30‑hectare experimental farm near Yamoussoukro uses advanced plant‑science methods to produce high‑yield, disease‑resistant cocoa plantlets. This work is further supported by the newly inaugurated Nestlé Institute of Agricultural Sciences in Lausanne, Switzerland, which leverages its global plant‑science expertise (including research from Nestlé’s Plant Science Centre in Tours, France) to develop low‑carbon, high‑quality cocoa, coffee and other crops—and then scale those innovations across farm networks.

Crop disease detection

How might we enable scalable, low-cost, early detection of Cocoa Swollen Shoot Virus (CSSV) to protect yields, improve farmer livelihoods, and stabilise cocoa supply chains?

Cocoa Swollen Shoot Virus (CSSV) is a major and growing threat to cocoa farming in West Africa. It spreads silently, often infecting trees for months before visible symptoms appear, and by the time signs like leaf reddening, shoot swelling, or dieback are detectable, it has already spread beyond control. Current diagnostic methods are slow, expensive, and impractical at scale, with error margins as high as 20%. These limitations hinder early intervention and make coordinated disease management across the region nearly impossible.

As a result, Cocoa prices have spiked to more than $10,000 USD per tonne – a 400% increase – and cocoa production is projected to fall by over 20% this season. This instability affects not just farmers and traders, but the entire chocolate value chain, from major food companies to the consumers now experiencing shrinkflation at the supermarket.

With this challenge, Nestlé is seeking AI/ML solutions that can detect CSSV at early or even pre-symptomatic stages, and that one can operate effectively in low-infrastructure, field-based environments.

Promising approaches may offer:

• Image or spectral analysis using handheld or drone-based devices to detect subtle plant health changes.
• Predictive modelling which integrates plant, environmental, and/or spatial data to identify at-risk zones before symptoms emerge.
• Cloud-based systems that support region-wide disease monitoring and scalable data aggregation for real-time response.
• A lower-cost and more efficient approach than current tools, which can take one hour to test 8 samples, and cost CHF 15 per sample.

Scalable carbon monitoring in agroforestry

How might we accurately and assess at scale above-ground biomass in agroforestry systems, enabling transparent carbon reporting and accelerating regenerative transitions?

Nestlé’s transition to regenerative agriculture in West Africa involves supporting over 20 million newly planted trees across thousands of smallholder cocoa farms. These trees are vital to improving soil health, biodiversity, and carbon capture, potentially cutting cocoa farming emissions (about 30% of Nestlé’s total) to near zero within a decade. However, accurately measuring this impact is costly and difficult.

Measuring changes in above-ground biomass across such a fragmented and complex landscape is extremely difficult. Existing methods often involve physical field measurements or costly high-resolution satellite tasking, which together can account for nearly 80% of total project costs. Accuracy remains below 50%, hampering transparency, slowing progress, and weakening stakeholder confidence.

This challenge calls for innovative AI/ML solutions that can transform how agroforestry carbon benefits are measured and reported. A successful solution could help unlock large-scale climate finance for farmers, provide proof of progress to buyers and regulators, and create a replicable model for sustainable agriculture measurement worldwide. Applicants with capabilities in AI for geospatial analysis, satellite or drone data interpretation, and carbon accounting will be well placed to contribute.

Promising directions include:

• Models that can estimate biomass using low-cost satellite imagery or drone-collected data, across variable canopy structures, terrain types, and planting schemes (typical of smallholder cocoa farms).
• Scalable solutions capable of aggregating data from thousands of dispersed farms into meaningful carbon reporting insights.
• Systems that integrate existing agroforestry or farm registry datasets to improve model performance and verification.

Dale Farm is the largest UK farmer-owned dairy cooperative, headquartered in Belfast, with over 1,000 members across the UK. Dale Farm is an end-to-end business, from farm to dairy manufacture to commercial sales. Dale Farm also owns a feed business, United Feeds, who sell animal feed products to both Dale Farm members and non-members.

The innovation challenges set out are specific to Dale Farm’s base in Northern Ireland.

Milk volume and composition forecasting

How might we accurately forecast future milk volumes and composition to inform long-term production planning and investment strategies?

Developing a robust business plan to support sustainable growth in the dairy industry requires a clear understanding of future milk supply. Reliable forecasting of both milk volume and composition (e.g. butter fat % and protein %) is critical for setting strategic goals, planning production, and anticipating costs. Sales records of milk volumes and historical milk composition data can uncover valuable patterns and trends, while external factors such as weather conditions, soil nutrients, feed type, and farming practices are also likely to influence milk supply.

Dale Farm is looking to develop a robust 5 to10 year business plan to support long-term production planning and future investment strategies. As milk supply is critical to their continued growth, gaining a clear understanding of future supply trends will be essential in shaping those plans.

This challenge seeks AI/ML solutions that develop data-driven forecasting models integrating historical milk supply and composition data with variable external factors to predict future milk volumes and composition. These models should be capable of identifying key predictive variables and generating actionable insights to support strategic decision-making for long-term production planning, investment strategies, and cost forecasting.

Ideal solutions will:

• Integrate available historical volume and compositional data with that of various external factors including farm parlour type, livestock housing type, feed type, milk pricing, bonus incentive schemes, and/or regional differences.
• Leverage wider datasets to improve model accuracy (e.g., weather conditions or soil nutrient levels).
• Enable predictive capabilities over industry-appropriate, and therefore actionable, timescales.
• Aggregate data into meaningful and actionable reporting insights.

Data-driven nutrition for dairy optimisation

How might we use milk production and composition data to develop predictive models that generate automated, optimised feed recommendations for dairy farmers?

Farmers often source customised animal feed to achieve specific on-farm goals, such as increasing milk yield or improving butterfat and protein levels. Feed recipes are currently developed through an experience-based, iterative process, relying heavily on farmer feedback to assess whether nutritional targets are met. However, with access to milk collection data (volume and composition) and feed purchase records, there is an opportunity to leverage data science to better understand the relationship between feed input and milk output.

Dale Farm is looking for a predictive model that can generate automated feed recommendations tailored to individual farm goals. In parallel, the model should assess the carbon footprint of different feed recipes, maximising both farmer returns and operational efficiency for the feed business.

This challenge invites AI/ML proposals to develop a data-driven solution linking each milk collection to feed types and feeding strategies using available lab and flow meter data from Dale Farm’s Just In Time production plant, United Feeds.

Ideal solutions will:

• Ingest and link data from multiple sources (milk collection records, lab analysis, feed purchase records, and production plant flow meters) to generate optimised, farm-specific feed mix suggestions aligned with individual goals.
• Estimate and compare the environmental impact (e.g., carbon footprint) of different feed recipes.
• Enable continuous refinement recommendations using feedback loops from updated milk output and feed data.

If the industry partner-led challenges do not align with your current project goals, we also welcome proposals that address critical issues aligned with one of the four challenge owners and their respective agrifood subsectors. However, we strongly recommend applicants consider applying to the challenges proposed by the industry partners, as these offer more direct alignment with their priorities and areas of potential impact.

Your project must:

• Demonstrate an innovative and ambitious idea that is technically feasible and scalable.
• Utilise AI/ML technologies or enable their adoption to deliver measurable improvements in business productivity, supporting the growth of the UK’s agricultural and food industries.
• Identify and respond to key challenges faced by the industry partner and across the agrifood value chain, highlighting how your solution provides practical benefits to them and the broader agrifood sector.