Everyday AI: reflections on the artificial intelligence in agriculture forum

This article captures key reflections from the artificial intelligence (AI) in agriculture forum, delivered as part of CyNam and HCR Law’s Everyday AI series. The session brought together farmers, technologists, legal experts, and researchers to explore the use of AI across the agricultural sector. The discussion focused on what AI looks like in practice today and the barriers that still need to be addressed to ensure its responsible and effective adoption. This write-up summarises the main themes that emerged from the conversation and connects them to wider trends shaping AI use in UK agriculture.

What is artificial intelligence?

One of the key themes from the forum was the debate around what qualifies as Artificial Intelligence. Participants raised questions like “Where is the line between existing technology and AI?” and “Am I already using AI without realising it?” To help anchor the following reflections and build a shared understanding among participants and wider stakeholders, we have included definitions of key terms.

We draw on existing frameworks to explain the different types of AI by capability and subsets of existing capabilities. We have utilised these definitions to map the use cases shared in the forum, showcasing examples of how AI is being adopted in the sector.

Categorising AI by capabilities

Artificial Narrow Intelligence: This is the only type of AI that currently exists. It can be trained to perform a single or narrow task, and is often used because it can accomplish that task faster and/or better than a human can. This may also be referred to as Weak AI.

Artificial General Intelligence: This is a theoretical concept. It can use previous learnings and skills to complete new tasks in a different context without human intervention for training purposes. Theoretically, AGI could learn and perform any intellectual task that a human can. This may also be referred to as Strong AI.

Super AI: This is also a theoretical concept. If it were ever made a reality, it would have cognitive abilities that surpass those of humans. It would be able to feel emotions, have needs and possess beliefs and desires of its own.

Subsets of Artificial Intelligence

Machine Learning: This is a subset of AI that focuses on creating models and algorithms that enable computers to learn from data autonomously, thereby becoming more efficient and effective.

In farming, machine learning allows systems to learn from past datasets, including yield information, animal health benchmarks, and environmental metrics, to make informed forecasts or generate alerts. For example, a health monitoring collar on a cow could learn its normal movement and feeding patterns, and identify subtle deviations that may indicate illness before a human detects it.

Machine learning can be divided further, and common subsets include:

• Supervised Learning: The input data and corresponding output labels are paired, and the machine learning model is trained on the labeled data. It discovers patterns from the labeled data, and can then predict or categorise new data.
• Unsupervised Learning: This involves training the machine learning model on data that is not labeled. The model still identifies patterns in the data, such as anomaly detection.
• Reinforcement Learning: The model is trained using a process of trial and error, with the model getting feedback in the form of rewards or punishments depending on its actions.

Deep Learning: This is a subset of machine learning (which is a subset of AI) that focuses on creating artificial neural networks, inspired by the human brain, that can learn from data and predict or decide a course of action based solely on that data. Therefore, whilst machine learning models depend on structured data and human-defined features, deep learning models use artificial neural networks to learn from raw data automatically. It is particularly effective in processing unstructured data like video footage or drone imagery. On a farm, this could be used to identify crop diseases, detect animal lameness or classify the ripeness of fruit automatically.

Natural Language Processing (NLP): This is a subset of AI that enables computers to comprehend, interpret, and produce human language. This is less common in agriculture, but could be used for voice-enabled farm assistants (like Siri for farming). Farmers could ask for real-time feeding schedules, sensor updates etc. using simple voice commands.

Generative AI: This is AI that can create original content, such as text, images, videos, audio, or software code, in response to a user’s request. This relies on deep learning models. Whilst NLP interprets and processes human language, generative AI creates new content by learning from patterns in data.

Computer Vision: This is a subset of AI that enables machines to see and interpret visual information. On a farm, this could include detecting pests, monitoring poultry behaviour and counting livestock.

Expert Systems: This is a subset of AI which simulates the decision-making capabilities of a human expert in a particular area. A repository of knowledge is created by a domain expert, which is then processed by an inference engine. A user interacts with the system through a user interface. In agriculture, these systems could be used to help diagnose crop diseases, or guide pest management strategies, especially when access to agronomists is limited.

Multiple AI models can be combined to create a compound AI system to solve complex problems more effectively than a single AI model could. For example, a customer service chatbot combines NLP models to analyse customer inquiries, and then is powered by generative AI to engage in conversation and suggest solutions to the customer.

Each subset of AI contributes something different, including learning from data, recognising patterns, understanding human language, interpreting images and more.

How is AI currently being used in agriculture?

To bring the above definitions to life, this section connects the capabilities outlined above to real-world use cases that were highlighted by participants and discussed during the session. While not exhaustive, it shows that Artificial Intelligence is already supporting a variety of agricultural tasks across livestock, crops, infrastructure, and compliance.

Predictive Health Analytics for Livestock

One of the uses of machine learning in agriculture is livestock health monitoring. Companies have developed wearable sensors that track each cow’s temperature, movement, eating habits, and rumination patterns. These systems analyse thousands of data points using machine learning to identify deviations from the norm. The farmer receives alerts through a dashboard or app which highlight potential issues, often hours or days before visible symptoms appear. This early detection not only improves animal welfare but also prevents costly production losses by allowing timely treatment.

Automated Lameness Detection in Cattle

A demonstration of deep learning and computer vision in agriculture is the use of overhead video capture combined with AI-driven gait and behaviour analysis to automatically identify early indicators of lameness in dairy herds. Using standard cameras installed in milking parlours, deep learning models interpret mobility patterns, detect abnormalities, and autonomously flag animals that may require attention. Again, these systems can identify issues earlier and with greater consistency than manual observation, supporting improved herd health management, reducing associated costs, and lowering environmental impact.

AI Integration in Robotic Milking Systems

AI-enabled robotic milking systems integrate advanced automation with computer vision, machine learning, and sensor analytics to optimise dairy production and animal welfare. A computer vision system recognises each cow and guides robotic arms for rapid and accurate teat attachment. Throughout the milking process, sensors capture data on yield, flow rate, behaviour, and indicators of udder health. This data stream is processed by machine learning models that detect trends, flag early signs of issues such as mastitis or declining performance, and support more informed herd-management decisions.

Precision Weed Control

Enabled by advances in computer vision and machine learning, precision-spraying systems can allow more efficient and targeted herbicide application. High-resolution cameras mounted on sprayers capture real-time field imagery, while machine learning models analyse colour, shape, and other visual features to distinguish weeds from crops or bare ground. When a weed is detected, the system activates individual nozzles to apply herbicide only where necessary. This fusion of computer vision, machine learning and robotics enables plant-level treatment, reducing chemical use, lowering costs, and improving overall input efficiency.

Poultry Welfare and Environmental Monitoring

AI-driven monitoring in poultry farming uses integrated environmental sensing and computer vision to provide insight into flock behaviour and health. Networks of sensors capture temperature, humidity, light levels, and patterns in feed and water intake, while camera feeds supply real-time visual data on movement and activity. Machine-learning models analyse these data points to detect anomalies such as crowding, reduced mobility, or abnormal clustering. In one application, vision-based algorithms identified heat-stress-related grouping early, enabling environmental adjustments that prevented losses. This technology enhances welfare management, particularly in high-density operations where human observation cannot maintain constant oversight.

Regulatory Compliance

Artificial intelligence can help farmers meet regulatory obligations by automating complex data analysis and improving decision-making at the field level. Tools powered by machine learning can process a farm’s soil data, land boundaries, and environmental designations, such as Nitrate Vulnerable Zones or Sites of Special Scientific Interest, to create specific nutrient and manure management plans that align with UK rules. These systems can integrate multiple spatial data layers, including slope, proximity to watercourses, and soil type, to flag high-risk fields and generate compliant application maps.

On the ground, mobile farm management tools are increasingly using AI to analyse input records (such as fertiliser or pesticide use), identify missing or incorrect data, and provide alerts where actions may breach regulatory thresholds. More advanced systems using computer vision are being trialled to identify specific pests on crops, reducing chemical use, overhead costs, and supporting compliance with pesticide regulations.

Environmental Reporting

Environmental data collection and reporting are becoming increasingly important in relation to net-zero commitments, environmental management plans, and biodiversity monitoring requirements. AI technologies can support these needs by automating the collection, processing, and interpretation of complex datasets.

For instance, machine learning models can be used to estimate greenhouse gas emissions from farm operations by analysing structured data on livestock, fertiliser use, and cropping systems, generating emission profiles aligned with national reporting standards. Remote sensing data can be processed using deep learning models to assess vegetation indices, soil characteristics, and indicators of land-use change, supporting carbon accounting and land condition reporting.

In the context of biodiversity, acoustic sensors and trail cameras equipped with computer vision can identify species from sounds or images, providing a scalable method for tracking pollinators, birds, or mammals. These technologies can support environmental reporting and reduce the manual effort associated with evidencing outcomes across a range of land management activities.

The Challenge of Interoperability

A reflection from the forum was the challenge of interoperability across the digital tools now used on farms. Participants noted that while AI-driven systems are becoming more common, many still function as closed platforms without open APIs (Application Programming Interface), making it difficult to move or combine data between them. A farmer described a situation where they were using multiple apps only to find that none of them communicated with each other. This lack of integration often resulted in duplicated effort and limited the value that could be extracted from the data being collected.

There was also recognition that some newer technologies are beginning to offer greater connectivity, but this varies widely across the market. As discussed in the session, considerations such as ease of integration, long‑term compatibility, and the ability to link with existing farm systems are becoming just as important as cost, simplicity, and performance. Overall, it can be deduced that improving interoperability will be essential if AI is to support more joined-up, data-informed decision-making across farming operations.

Building Trust, Ensuring Value, and Protecting Data

As AI tools become more common in the agricultural sector, forum participants consistently raised three practical concerns: trust, cost, and data ownership. Trust depends not only on how accurate a system is, but also on how transparent it is. Farmers need to understand how AI systems reach their conclusions to act on them with confidence. Cost remains a key factor, particularly for small and medium-sized farms where return on investment must be clear. Alongside this, questions around data ownership remain. With so much sensitive information now collected, farmers need clarity over how their data is stored, accessed, and used by third parties.

These considerations are as important as any technical features. For AI to deliver value in agriculture, it must be explainable, cost-effective, and have data protection at its core.

Conclusion

The discussion highlighted that while AI has already found meaningful applications in agriculture, adoption remains uneven. Participants made it clear that more support is needed to navigate these technologies and address the concerns of farmers around trust, cost and data privacy. This will be key to unlocking the sector-wide benefits of AI. As the technology evolves, continued dialogue between farmers, developers, legal experts, and researchers will be essential to ensure tech solutions best answer the real-world needs of the agriculture sector.