Whether it’s autonomous orchard sprayers, software solutions that assess data or sorting machines helping packers and shippers screen for imperfect fruits and vegetables, artificial intelligence is taking hold throughout the fresh produce supply chain.
And it doesn’t show any signs of slowing down.
In fact, Yiannis Ampatzidis, a University of Florida’s Institute of Food and Agricultural Sciences associate professor of agricultural and biological engineering, calls AI the fourth agri-revolution. Ampatzidis, who studies precision agriculture and mechanization for crop production, has seen firsthand the importance — and potential — of AI.
Where things stand with AI
When he first joined UF/IFAS, Ampatzidis says growers needed a way to assess damage in citrus groves and on farms following hurricanes for reporting to the USDA for crop insurance policies.
This information initially came in the form of drone imagery, but as technology improved, AI could take what the drone saw and begin processing the data in a way that helps growers get a better idea of what’s happening in their farms and groves.
With the devastation from huanglongbing, also known as HLB or citrus greening disease, growers wanted a more accurate tree inventory using drone imagery. Ampatzidis says UF/IFAS developed Agroview, a cloud- and AI-based technology to process, analyze and visualize data collected from unmanned aerial vehicles like drones, as well as other sources such as satellites, smart sprayers and more.
“We developed this aggregate technology that utilizes a lot of smart algorithms and AI,” he says. “We can detect the trees, measure the tree volume and even measure the tree stress. Doing that, we can reduce the data collection and visualization time maybe even up to 90%.”
Ampatzidis says the next step will be to take the multispectral images that the drone captures to assess plants’ fertility needs, a critical step in combating citrus greening. Crop insurance companies also use this technology to help growers assess how many trees to replace from citrus greening or after a hurricane, he says.
Another use is with smart sprayers, mainly used in citrus groves, which use LIDAR and cameras to sense and assess trees and tree canopies and target nozzles to deliver an appropriate coverage for the size. Ampatzidis says this system also detects other objects, such as poles, irrigation systems, workers and more.
These cameras also collect imagery and data, Ampatzidis says, and his research team uses what the smart sprayer collects to detect fruit density and count fruit; in the future, it could help growers better predict yield and fruit drop.
“In Florida, we did some studies, and we found out that you can save 30% of the chemicals using these technologies,” he says.
Sunkist says it deploys AI in its packinghouses through its Sunsortai; using machine learning, its computers use data and algorithms to make predictions on fruit sorting and sizing. Sunsortai scans each fruit for diseases, misshapenness, decay, blemishes, weight, blemishes and more.
“As we provide more images over time, technology can learn and adapt to identify new capabilities, operate at a higher efficiency and enhance quality control for improvements and defects in fruits,” says Mina Abdelshahid, director of operations for Sunkist RTS Operations. “Our system can also program names for common defects, diseases and other imperfections into our database to make them more easily identifiable and easier to use in the future.”
Challenges for AI adoption
A big obstacle to widescale adoption of AI on the farm is the farm itself, says Fiona Turner, CEO of Bitwise, an Australian startup using AI to provide precision analytics.
“No two farms are the same, so it takes an initial time investment to get the systems set up and working effectively,” she says. “Farmers are busy, so they can be tempted to keep doing things how they’ve always done them — but the world is changing, the weather is changing and farms need to adapt with it.”
Training is a key barrier to adoption in the ag industry, says Elliott Grant, CEO of Mineral, an independent sustainable agriculture company under Alphabet (the parent company of Google). AI technology providers need to help growers understand how to best use the data these systems collect.
“We’re optimistic that with more education on the opportunities AI can bring, the efficiencies that can be found, and the focus on putting [good] data to work, that this barrier will be much lower,” he says.
Grant says another challenge is that farmers might not see the value in adopting AI-based technology. It might be easy to see the return on investment for a new tractor or commodity, but with technology the ROI is often found in improvements in efficiency, which can be harder to quantify.
“When AI models are robust and reliable on a farmer’s field, and create a clear ROI, they will be widely adopted,” he says. “But we have a way to go before we get there.”
Another consideration, Turner says, is that while AI helps speed up processes, it isn’t infallible.
“There is always a risk of errors or inaccuracies in AI predictions, particularly when dealing with complex and dynamic agricultural systems,” she says. “The best technology partnerships don’t replace the farmer’s knowledge; they add to it.”
What does the future hold?
Grant points to autonomous and robotic weeders, yield forecasting models and AI-powered fruit and produce sorters as other ways that AI has infiltrated and improved production.
But he says it’s important for those in the produce industry to see AI for what it is: as a productivity tool.
“Some of the really compelling applications we’re seeing [are] in plant identification, yield forecasting and quality inspection,” he says. “We believe that in the future, generative AI agents will become the interface through which farmers access information, query datasets and gain new insight — replacing what farmers are working with today whether it’s clunky database searches, inscrutable graphs or colorful but hard-to-interpret field maps.”
Turner agrees. “We’ve only scratched the surface with how AI can help farmers.”
Grant says it’s a challenge to predict exactly where AI will take the fresh produce industry in the next decade, given how quickly things change.
“AI could be 100 to 1,000 times more powerful in five to 10 years, given the exponential trajectory that we’ve seen to date,” he says. “It’s reasonable to expect that automation and the AI tools — from plant identification to yield forecasting to quality inspection — will become commonplace within 10 years, especially given the intense labor availability pressure that specialty growers face.”
Grant says the next frontier for AI will be in data entry, which could take the form of pre-filling fields, context-specific questions and checking responses for accuracy.
“Replacing manual data entry with AI tools will be able to save farmers time, improve accuracy and give them the mindshare to focus their attention on more meaningful on-farm tasks,” he says. “We are in the early days of seeing AI be incorporated into workflows, jobs and equipment. Bear in mind it took more than 30 years for electricity to replace steam power … it takes time for systems and behaviors to fully adapt.”
Turner says she expects AI implementation to continue growing on farms in the next five to 10 years and to improve efficiency and operations.
“Predictive analytics is likely to be huge for better understanding weather patterns, pest outbreaks and market demand,” she says. “We expect to see AI-powered tools that enable precision agriculture at a more granular level, allowing growers to monitor and manage their crops with unprecedented accuracy and in real-time. Predictive analytics will enable growers to anticipate and mitigate potential risks and challenges more effectively through analysis of historical data, environmental factors and market trends.”
Grant says he also sees AI having a potential impact on seed breeding in the future, as growers could better match future crops and varieties to local conditions. He says AI could also help speed up the discovery of molecules and biologicals for new pesticides, herbicides and fungicides with less impact on soil health and biodiversity. AI has the potential to help develop better financial and risk management tools for growers, he says.
It’s not just for produce growers
The yield-data growers collect has the potential to have a bigger impact on the supply chain, says Ampatzidis of the UF/IFAS. As the season progresses, growers will better inform packinghouses on what to expect in terms of yield and timing of harvest.
“It can really improve and optimize the logistics and trading of the products and the commodities,” he says.
Turner says AI will help packers and shippers reduce food waste through optimization.
“There is potential for AI to also support the supply chain by optimizing transportation routes, scheduling and logistics to ensure timely delivery of fresh produce to retailers and consumers,” she says. “This can help reduce food spoilage, increase shelf life and enhance overall supply chain resilience.”
AI, too, will help retailers better promote what’s new and what’s in season, Turner says.
“We’ve also already started to see AI’s use in personalized marketing and consumer insights, analyzing consumer preferences, purchasing behavior and market trends to provide insights for product development, marketing and reducing waste,” she says.
Take-home message
Ampatzidis says while it’s a common fear that AI will replace all workers in almost every situation, that is untrue.
“Some people worry that AI will take our jobs, but we are not at that stage right now,” he says.
But this is not a time to ignore AI, either, he says. Now is a good time — when there’s less immediate need to implement these new technologies — to begin to understand how AI and new technology could improve an operation.
“You really need to educate yourself and try [to] at least be aware of what’s happening, what [these technologies] can offer,” he says.
Grant says it’s good to start small when introducing new technology, especially with AI.
“Find ‘baby step’ applications to get started and familiar with AI,” he says. “Like any new technology, adoption will require familiarity and behavior change. Starting with a couple of small projects — such as a chatbot interface to an internal dataset for salespeople or an AI-summarization tool for written reports or emails — will build the company’s capabilities and awareness.”
Grant says it’s a change of mindset, too.
“Rather than think about AI replacing existing tasks, consider ways that AI could change the way work is done instead,” he says. “It needs to be implemented as an augmentation tool, not a replacement. But the efficiency, speed, consistency, and power of AI can unlock new ways of working.”