.png)
John Deere
Agriculture · US
50% less herbicide
John Deere's See and Spray uses computer vision to spray only the weed, not the field
See and Spray integrates physical farming equipment with real-time computer vision and nozzle-level control. It is not a standalone AI tool. It is a detect, decide, and spray workflow embedded in the machine itself.
Friction
Farmers apply herbicide across entire fields, including areas with no weeds. With rising input costs, sustainability pressure, and herbicide resistance growing, broad-spectrum spraying was becoming both economically and environmentally untenable.
Breakthrough
Boom-mounted cameras scan the field continuously. Onboard processors and AI models identify weeds in milliseconds. The ExactApply system then activates only the required nozzles. In crop applications, the system distinguishes crop from weed; on fallow fields, it detects any green plant on bare soil.
Impact
Used on 5+ million acres in the 2025 growing season. Average of nearly 50% less non-residual herbicide use. Nearly 31 million gallons of herbicide mix saved. An Iowa State University demonstration on 415 soybean acres showed 76% product savings and nearly $6,500 in herbicide cost reduction.
Unlock the full analysis with breakthrough, impact, what made it smart and its technical approach below!
Problem
Traditional herbicide application is a blunt instrument: spray everything, whether there is a weed or not. With herbicide costs rising and resistance increasing, farmers needed a smarter approach that could work at field scale and machine speed.
What made it smart
The combination of hardware and software in a single integrated machine is what makes this work. AI does not just advise, it directly controls physical nozzles in real time, at up to 15 mph, scanning 2,500 square feet per second.
Technical approach
High-resolution cameras feed onboard processors that run AI models for weed detection and classification. The ExactApply nozzle control system receives signals from the AI and fires only where needed. The John Deere Operations Center provides as-applied maps and weed pressure data after each pass, creating a data feedback loop for future decisions.
Strategic lesson
The highest-value AI applications are not advisory dashboards. They are systems where AI directly controls a physical action at the moment of decision.
Reflection question
Where in your operations does a human currently execute a rule-based physical action based on information that AI could process faster and more accurately?