While Artificial Intelligence (AI) promises to improve agricultural sustainability, the technology itself carries a substantial carbon and water footprint the Institution of Agricultural Engineers (IAgrE) AI conference heard.
Fernando Auat Cheein, an international expert on robotics from Harper Adams University raised concerns about the energy consumption required to train and run AI systems, noting that sustainable AI practices were essential for the technology to truly benefit agriculture. He highlighted a Royal Academy of Engineering report on sustainable AI, which recognised this growing concern.
READ MORE: Baroness Minette Batters' Farm Profitability Review: Industry reacts as review published
Yet speakers also highlighted AI's potential to reduce agriculture's overall environmental impact.
Professor Simon Pearson, Lincoln University, who holds the IAgrE Award of Merit pointed to improvements in nitrogen use efficiency as a prime example, where AI-driven precision could significantly reduce both costs and carbon emissions. Kieran Fitzgerald, vice-president, digital services at DeLaval shared how early disease detection in dairy herds mitigates production losses, reducing the environmental cost per unit of milk produced.
The conference explored how AI supports regenerative farming practices.
Jonathan Henry, managing director of Garford Farm Machinery emphasised the complementary relationship, with AI enabling detailed soil health analysis that underpins regenerative approaches.
Prof Pearson noted that AI can help navigate the uncertainty in regenerative farming, accelerating system improvements and providing traceability evidence increasingly demanded by consumers and regulators.
One delegate reminded participants of a fundamental truth: "AI is a machine, not a being."
This perspective helps frame the technology as a tool that must justify its environmental cost through measurable improvements in agricultural efficiency and sustainability, the conference heard.
What happens when AI-powered farms lose internet access
The vulnerability of cloud-dependent AI systems became a critical discussion point when Mr Fitzgerald revealed that a Roblox-related AWS (Amazon World Services) hack had taken down DeLaval's entire system.
"If you have business-critical services in the cloud, you will fail at some point," Mr Fitzgerald warned delegates.
The incident demonstrated the risks facing farms that depend entirely on cloud-based AI for daily operations, from automated milking systems to decision support platforms.
Morten Bilde, managing director at AGCO said that relying on cloud computing for multiple solutions and integrations in the technology stack created significant challenges, particularly given agriculture's often-limited connectivity in rural areas.
The solution, according to multiple speakers, lies in edge computing - processing data locally on farm equipment rather than depending on constant internet connectivity.
However, implementing edge computing brings its own challenges.
Systems must remain reliable in harsh environments with dust, moisture, temperature extremes, and mechanical vibration.
Allan Kildeby, head of camera sensors at CLAAS noted that hardware accelerators and localised processing capabilities were becoming essential building blocks for agricultural AI systems.
The mixed fleet strategy discussed by AGCO offers another approach to resilience. Rather than creating closed ecosystems dependent on single providers, designing solutions that integrate with various machines and systems reduces the risk of total system failure. Their forest harvester solution, developed as an aftermarket kit, exemplifies this approach.
Mr Henry stressed the importance of rigorous testing and verification to ensure AI systems remained reliable under real-world farming conditions, where internet connectivity cannot be guaranteed.
READ MORE: Environmental reps walk away from Scottish Gov's 'failed' farming policy
