Researcher Wins Agricultural Competition Using AI to Guide Crop Decisions
Nipuna Chamara, a research assistant professor in biological systems engineering at the University of Nebraska-Lincoln, won the highest corn yield category at the 2025 Testing Ag Performance Solutions competition by using artificial intelligence to make decisions about when to plant, fertilize, and manage his crops.
The result suggests that AI can help farmers optimize their operations when combined with human expertise. "If a person like me, who's not a farmer, can use AI to win a competition like this, imagine what a seasoned farmer, with decades of experience and knowledge, could do with this tool," Chamara said.
How the Competition Works
TAPS is a university-led competition where participants manage actual plots of corn and soybeans for a full growing season. Teams make real-world decisions-seed selection, irrigation, pest control, grain marketing-in a low-risk environment where they can experiment without affecting their home operations.
Extension educators provide teams with data on soil moisture and health. Teams then decide when and how much to water, fertilize, or take other actions. The 2025 competition included 116 teams managing plots at research centers in North Platte and Mead, Nebraska.
Chris Proctor, a Nebraska Extension educator managing the contest, noted the value of this structure. "The average grower has about 40 growing seasons to improve their operation," he said. "Within TAPS, last year we got 116 teams competing, so that's 116 growing seasons' worth of decisions all in one."
Chamara's First Attempt: 2024
Chamara entered the competition in 2024 with support from his doctoral supervisor, Yufeng Ge, and faculty experts in nitrogen management, irrigation, agronomy, and agricultural economics. He used OpenAI's ChatGPT to help guide his decisions.
The process was manual and cumbersome. ChatGPT had no real-time data access, so Chamara's team uploaded information about crop type, soil health, moisture, and weather manually. They then asked the AI close-ended questions about planting and fertilizing timing.
Competition managers did not view the AI use as cheating. "At that time, the excitement was building around AI, and I still don't know that all of us really had our heads wrapped around what it was," Proctor said. Chamara's team placed seventh in the yield category that year.
Significant Improvements in 2025
By 2025, AI capabilities had advanced noticeably. The systems could pull real-time data from the internet and had access to data from the previous year's competition, providing context for new recommendations.
Chamara grew three corn plots and one soybean plot, uploading weekly data like grain reports and asking focused questions with specific objectives. The AI would combine internet data with his historical and current information, then recommend actions with explanations.
One example: the AI suggested Chamara lock in corn prices early because market prices were fluctuating due to new tariffs. This recommendation came from the AI analyzing recent news about commodity price movements.
Limitations and Safeguards
Chamara acknowledged real risks. AI can base recommendations on faulty or incorrect information found online. He stressed the importance of verifying recommendations against reliable sources like grain reports and extension publications.
The United States and Canada's long history of collecting and sharing agricultural data openly helps AI systems make better decisions, he noted. Farmers with robust digital records can train AI to be more useful by providing context and historical reference points.
What's Next
Chamara and his team published their research outcomes in the journal Artificial Intelligence in Agriculture. For the 2026 competition, he plans to focus on profitability and sustainability.
He also wants to develop an app that lets farmers connect AI directly to farm sensors, automatically uploading real-time data daily. This would give AI the most current information to guide decisions.
Chamara said farmers already possess the domain knowledge needed to work effectively with AI. "Like a person who has Google, or a person who uses the library, we can use AI as a tool to make us more powerful in processing data," he said. The key is learning to ask the right questions through effective prompt engineering.
Your membership also unlocks:
