Agriculture is at the nexus of a host of environmental challenges, from the need to feed the world and provide people with livelihoods to land and water stewardship, pollution reduction, biodiversity protection, and climate adaptation.
This is not an easy feat. Farmers have to grow more resilient food supplies in the face of increasingly unpredictable weather, land degradation, input costs, water scarcity, market demands, and shifting consumer preferences. Meanwhile, public and private leaders must develop agricultural systems that enhance production while not compromising the natural systems that production depends on.
This is why AI and digital technologies in agriculture have attracted attention.
AI has the potential to assist farmers and governments in improving the information they need to make informed decisions about crop monitoring, climate forecasting, soil and pest management, water use and irrigation, biodiversity monitoring and supply chain efficiency, and where inputs are being used inefficiently or where the land is stressed.
However, AI cannot just be assumed to be a good thing for agriculture, nature or climate.
Digital agriculture can promote positive outcomes, but it also has the potential to exacerbate existing negative trends and problems, especially if it is designed around productivity targets alone, if it takes a proprietary platform approach, if it does not include data governance or rights for farmers, and if the technologies are too expensive and complex for farmers to use.
The question to ask is not whether AI could make agriculture more productive. The question is whether it could make it more sustainable, more equitable, and more nature-positive.
Better information is essential for agriculture
Agriculture has long relied on information.
Every day, farmers make decisions based on weather, soil, pests, water, crop health, markets, animals, knowledge, traditions, and more. Farmers gather this information from their own observations and experiences and from others, ranging from family and friends to scientists and public agencies.
AI can be used to gather and provide even more information.
Information about weather, soil, water, land, climate, and crops can be collected and shared to help farmers identify issues before they become serious problems. Information can help a farmer detect crop stress early; help a water manager predict when farmers will need irrigation; and help a government agency predict drought or land-use changes.
This new information has the potential to improve the decisions made by farmers and other agricultural stakeholders. But information alone does not guarantee better decisions. If a farmer does not understand an AI tool’s output, does not trust that output, or is not in a position to make use of it, they probably will not make decisions that improve the health of their farm.
AI tools that provide data must also be interpretable, understandable, trustworthy, and accessible by farmers.
Use fewer unnecessary inputs
One of the most promising uses of AI in agriculture is to identify when and where farmers use inputs.
Nutrient inputs, such as pesticides, fertilizers, and herbicides, as well as other essential resources like energy and water, cost money to acquire, but they can also have adverse environmental impacts if used improperly or inefficiently.
When nutrients run off agricultural land, they can flow into rivers and lakes and damage the surrounding ecosystem. When pesticides are not applied properly, they may not only be ineffective, but they can also harm insects, pollinators and aquatic wildlife that are not pests. Water use that is not matched to crop demand puts strain on freshwater systems.
AI can help identify where and when to apply inputs and when they can be safely skipped entirely. It can support variable-rate applications, pest forecasts, soil analysis, crop health monitoring, and irrigation scheduling.
Reducing unnecessary use of inputs has the potential to help both farmers and environmental health.
Reducing environmental threats from pesticides, for example, is not only about using less hazardous chemicals and enforcing stricter regulations. It also involves determining when pest management is necessary, identifying the areas where pest risks are high, and finding ways to help farmers minimise exposure risks to pesticides.
Artificial intelligence could play a role in this transition, but only if it is implemented with a focus on risk reduction rather than the maximisation of output.
For example, digital advice that recommends increased inputs because this is commercially good for one supplier may make the problem worse. Tools that help reduce unnecessary input use while maintaining yields could play a positive role, both in farm resilience and the environment.
Water management is increasingly important
Water is one of the main challenges for agricultural producers. In some places, they face drought and water scarcity; others face flooding and variability in the timing and amount of rainfall. Often the two extremes appear within a year, as climate change makes historical weather patterns a less reliable guide.
AI could support this by integrating weather predictions, soil moisture data, crop water demands, and historical irrigation records to support farmers’ decisions on when to irrigate, how much to apply, and where problems are most likely to arise.
That is important because the way farmers use water impacts rivers, wetlands, aquifers, downstream users, and other parts of the catchment.
You cannot separate agriculture from freshwater protection. Decisions around irrigation could help to reduce demand on water systems, while land management decisions could help reduce runoff, soil erosion, and nutrient pollution.
However, digital tools to manage water use need governance, to avoid situations where AI allows some users to become more efficient at extracting water, yet the total water withdrawn from a catchment rises due to a lack of restrictions on total water use.
Efficiency at farm level does not automatically equate to sustainability at river basin level.
So, we should be using AI to help support the allocation, monitoring and planning of water, as well as just optimising on-farm decisions.
Supporting climate adaptation
Agriculture is already facing a changing climate.
From heat stress, to droughts, to floods, to new pest pressures, to shifts in growing seasons and extreme weather events, conditions for production are changing, and farmers need practical, affordable, locally relevant tools to adapt.
AI could support this through better seasonal forecasts, better crop vulnerability mapping, identification of better crop types, warnings of pest and disease outbreaks, and better planning of adaptation measures in the face of severe weather.
This may be particularly important for those who currently have lower access to more traditional extension and advisory services, where the use of mobile, voice-based and local language platforms could play a role in narrowing this gap if developed inclusively.
But climate adaptation is more than just a technical challenge. Farmers will also need greater access to credit, insurance, infrastructure, land tenure security, public advice and markets. AI may be able to support adaptation efforts, but it cannot, of itself, provide the wider policy context required to help farmers and other agricultural land managers adapt.
The goal of agriculture needs to be to meet our food needs while also achieving our biodiversity and climate goals, rather than simply improving short-term food production. That is the broader challenge behind balancing food production, biodiversity and climate goals.
Biodiversity must not be an afterthought
Agricultural production landscapes can either support or damage biodiversity.
Hedgerows and woodlands, wetlands, grasslands, field margins, trees in the field, soil, and watercourses can all provide habitat, while pollinators and predators can aid crop production, and healthy soils and landscape diversity can help farm resilience.
AI could be used to help map these features and monitor ecological change across individual farms and at broader landscape scales. It could support better targeting of funding towards restoration efforts and habitat creation, and the adoption of biodiversity-friendly farming practices. It could be used to identify where ecological corridors may be required, or where land management could be causing damage.
Yet biodiversity is not merely a matter of the readily measurable.
Certain species are elusive, some habitats require local know-how for full comprehension, and many ecological interactions are highly intricate. Consequently, if the sole focus of AI tools is restricted to the readily observable or quantifiable, the significance of those things may go unnoticed.
Furthermore, we need to consider whether the economic valuation of nature should be taken as such. While agriculture depends heavily on ecosystem services, nature should not only be valued insofar as it sustains production.
Artificial intelligence can assist in measuring and monitoring biodiversity, but the technology should enhance our ability to make ecological decisions, rather than taking that choice away from us.
Data rights and farmer autonomy
The agricultural deployment of AI is fundamentally built on the existence of data.
This might incorporate data on soil characteristics, output, equipment utilisation, input application, water consumption, livestock, field boundaries, accounts and market activities. This raises questions over data governance.
Important questions include:
- Who is the owner of the data?
- Who is permitted to sell it?
- Is it possible for farmers to transfer their data from platform to platform?
- Are the data capable of influencing insurance, credit, land values or access to markets?
- Might companies use these data in order to influence farmers towards their own products?
These are not trivial questions. Digital technologies can, after all, produce reliance.
Should farmers utilise proprietary platforms for recommendations, they can find themselves relinquishing power over decisions. Should the data be locked within a single platform, changing platforms will be more complicated. Should the application of AI tools prove costly, smaller farms may be excluded from their use.
Digital farming must enable rather than curtail the ability of farmers to act. For this reason, governments should establish rules which promote open standards, interoperability, transparent pricing and appropriate terms. Farmers must be empowered in terms of data and how it is used.
The danger of reduced farm variety
As AI systems are designed for the achievement of specific goals, we can expect that they will be trained towards the optimisation of such goals.
This will certainly have benefits, yet it might also be harmful. Inasmuch as an AI system is intended principally to maximise output or profit, it might overlook the value of biodiversity, soils, water, cultural traditions or long-term adaptability.
In addition, there is also a possibility of standardisation being promoted through these systems.
Agricultural systems vary across the globe because there are variations in climate, topography, geology and soils; cultural traditions also vary; markets differ. A piece of AI advice which might prove useful on one farm would be entirely inappropriate on another.
Local and Indigenous knowledge must not be replaced by automated decision-making systems.
The experiential knowledge of farmers is still essential. Artificial intelligence can serve as an aid to make decisions. It should not be the decision-maker. The most effective system will be an integration of data, local knowledge, public extension and ecological understanding.
Equity and access
The advantages offered through the application of AI in agriculture will be unequal unless policy ensures equitable access.
Larger farms might have a competitive advantage over small farms in terms of their ability to purchase sensors, drones, software and data services or pay for technical guidance, whereas small farms will have lower connectivity, lower investment power and less time for the trial of new technologies.
It may even deepen inequalities if digital agriculture serves already well-resourced farmers, rather than the rest.
Policymakers should be able to support the existence of open data, public extension systems, farmer cooperative platforms, training, rural internet access and low-cost services.
No farmer should be forced to choose between exclusion from digital agriculture or dependence on high-priced, private solutions.
The use of AI can lead to greater equity, but only if such a solution is embedded within policy design.
AI should support policymaking, not just farm management
While the conversation about AI and agriculture frequently centres on the individual farm, this technology holds significant value for shaping public policy.
Digital tools offer a way for governments to track land-use changes, review the impact of agri-environmental initiatives, spot early signs of water stress, direct restoration money effectively, analyse soil dangers, and gauge how well policies are functioning at a regional level.
All of these capabilities would contribute to stronger agricultural governance.
In practice, public authorities might leverage AI to pinpoint exactly where investment in soil recovery is critical, where the threat of water contamination is growing, or where habitat connectivity needs reinforcement. Furthermore, they could use these technologies to scrutinise whether the public funds being pumped into the sector are actually achieving the intended ecological results.
This matters because agricultural policy must do more than just pay farmers to do things. It must incentivise outcomes such as improved water quality, enhanced soil vitality, reduced greenhouse emissions, diminished pesticide hazards, and more robust biodiversity.
AI can help bring these outcomes into sharper focus.
The eco-impact of smart farming
Yet AI systems carry environmental costs too.
The sensors, drones, servers, data hubs, gadgets and network infrastructure required for digital farming consume energy and raw materials. Smart farming should not automatically be viewed as green just because it employs sophisticated technology.
The ecological price paid for using a specific tool must be commensurate with the ecological advantage it provides.
A basic advice service capable of lowering water consumption or pesticide danger would surely be worth it. But a highly sophisticated system, needing costly hardware for minor environmental gains, might not be.
So, while productivity gains are useful for gauging agricultural AI, the technology also needs to be evaluated against its actual environmental and social contribution.
Responsible AI in farming
Responsible AI in farming must adhere to these core principles:
- It must serve the farmer, not supplant their expertise.
- It must help reduce environmental harm as much as enhance yields.
- It must be cheap and reachable for everyone.
- It must safeguard the ownership of farm data.
- It must be honest about its advice and its limits of certainty.
- It must foster wildlife, soil wellbeing and watershed protection.
- It must work with regional contexts.
- It must reinforce public policy and extension services.
- It must be assessed against its own footprint.
These guidelines are essential because farming is not just another industry to be modernised. It is the foundation of food security, countryside livelihoods and countryside stewardship.
Digital technologies must serve a public goal
While AI can help agriculture meet the competing demands of feeding humanity, safeguarding nature and stabilising the climate, it can do this only in the right circumstances.
It will be valuable to the degree that it renders the impact of farming on the environment more transparent; helps people use water, fertiliser and pesticide with more care; promotes climate proofing and biodiversity measurement; and leads to better government planning. It will help governments work out where and how they need to act and check whether funding is really delivering the goods.
However, it will prove damaging to the extent that it increases reliance, marginalises small-scale holders, limits farmer autonomy, restricts agricultural diversity, or disregards the ecological price of intensification.
In other words, digital technologies must have a public objective.
Technology alone will not define farming’s future. Rather, it will be determined by effective policy, farmers’ knowledge, balanced markets, government investment, flourishing ecology and climate resilience.
AI can be part of that future, but only if deployed in ways that support such ends, rather than drawing attention from them.
Further reading
For further context on the environmental policy issues that inform digital agriculture, see:
