Weekly Update: AI is landing where farmers already work.
· Matt Anderson
Deere’s farm AI assistant, Google’s climate tools, and potato disease scanners that see before you do.
This week's AI/ML wrap for Australian agribusiness is less about startup splash and more about models and platforms landing where farmers already work. John Deere's new Operations Center assistant shows how generative AI is being bolted onto machinery telematics and paddock history, with data-control language that AU growers will scrutinise closely. At the same time, Google and the Gates Foundation are pouring serious money into satellite and AI climate and crop tools for smallholders, while research groups push disease detection earlier (soybeans in Illinois, seed potatoes in the Netherlands) and build multimodal models that can point to the exact leaf lesion, not just name it. For Australian readers, the through-line is decision support under climate and cost pressure: better sensing, better questions of your own data, and advisory systems grounded in trusted agronomic knowledge rather than generic chat.
1. Google and the Gates Foundation scale AI climate tools to 200 million smallholders (Climate & risk, 18–22 Sep)
$100m plus Google research support to expand AI micro-climate forecasting, sub-metre field mapping and crop land-use layers across Africa and South Asia. The AU link is that this is the same kind of satellite AI already used here for drought, water and production risk.
2. John Deere's "JD" assistant in Operations Center (Decision support, announced early Sep, still circulating)
Growers can ask plain-English questions of their own field, machine and ops data. Early access is US only, with a wider rollout flagged for later in 2026. It's relevant to AU Deere fleets once it gets here.
3. Croptimal Croptiscan 3000 (Labour & automation, 22 Sep)
A tractor-mounted camera boom trained to flag potato virus Y (PVY) and leafroll in seed potatoes before symptoms show. The company quotes about 1.5 ha/hour and EUR 95k. It fits AU seed-potato certification and the shortage of labour for hand-roguing out diseased plants.
4. SIU robot and AI for early soybean disease detection (Research, 24 Sep)
Southern Illinois University is building a ground robot that photographs leaf undersides, which drones miss, to train AI for early frogeye leaf spot detection. The aim is hotspot maps for site-specific fungicide. It's still research stage.
5. Oracle and Wild Bio on AI for crop resilience and carbon (Climate & risk, 24 Sep)
Oracle's cloud and field sensors capture trial data, and AI is applied to crop genetics for stress tolerance and auditable carbon uptake. These are partnership goals, not field results yet. It's relevant to AU wheat under heat and drought, and to carbon claims.
6. CABI releases AI-ready plant-health datasets on Hugging Face (Models & methods, 21 Sep)
Curated pest and disease data for building agronomy chatbots that are more reliable, released under the Gates-backed Generative AI for Agriculture project. A livestock dataset is due later this year.
7. AgriScope, a pixel-grounded multimodal AI for crop images (Research, 17 Sep, just outside the window)
A research paper on a model that marks exactly where on the image the disease, pest or weed is, rather than just naming it. It comes with a very large training dataset, and the results are reported by the authors.