Alliance Partner

Farmer-Centered AI in Agriculture: Making the Juice Worth the Squeeze

Producers, innovators and investors share insights on responsible artificial intelligence for agriculture.

August 7, 2026 | Dee Shore | 17-min. read

During a preconference tour, Chris Reberg-Horton, the N.C. Plant Sciences Initiative’s resilient agriculture platform director, showed off tractors rigged with rugged cameras designed to take the everyday wear and tear that takes place on farms.

Speaking to hundreds of innovators, investors and others about the future of artificial intelligence for farming, Smithfield Foods’ Kraig Westerbeek put producers’ needs in plain terms.

Westerbeek was among five producer panelists at the AI in Agriculture Conference at North Carolina State University in spring 2026. The event drew 460 growers, tech innovators, investors and researchers to explore applications in computer vision, robotics, connected devices, large language models and more.

Innovation must demonstrate clear economic value before it can be widely adopted.

Over three days, a clear consensus emerged among conference speakers: for AI to succeed on the farm, it must make economic sense, endure rugged conditions and keep farmers — not software — in control.

“Because agriculture operates under narrow economic margins, innovation must demonstrate clear economic value before it can be widely adopted,” said conference chair Daniela Jones, an NC State associate professor of biological and agricultural engineering and the N.C. Plant Sciences Initiative’s platform director for education and workforce development.

The Producers’ Perspectives

Despite differences in farm size and commodities, members of the growers’ panel echoed a unified message that set the conference tone: When AI is in use in agriculture, ultimate decision-making power must remain with the producer.

Westerbeek was insistent on this point. His division raises 11.7 million hogs annually for the world’s leading hog producer and processor. Smithfield Foods uses AI to help select sires and dams for specific growth characteristics and to streamline the movement of pigs to various barns throughout their lifespan.

Despite his support for AI-enabled tech, Westerbeek was clear that it cannot replace what he sees as a necessary human touch in livestock production.

“AI can be helpful in caring for animals, but ultimately … the relationship between the person caring for the animal and the animal has to be there,” he said.

Remaining Accountable to Customers

Another reason that producers need to remain front and center is because they are the ones accountable to their customers.

Nash Produce, a sweetpotato packer that sells to major retailers, uses AI-enabled cameras to visually inspect sweetpotatoes and help sort them by grade. This technology is intended to allow the facility to operate with fewer people, addressing the persistent challenge of finding and retaining labor in rural areas.

Company president Thomas Joyner said that while AI systems are “continually learning” and improving, they require constant attention and adjustment.

“When you’ve got issues that occur on our packing operations, when Walmart comes back and says, ‘Why’d you pack this?’ I can’t go back to the computer and say, ‘Why did you put this in the box?’ I’m going back to the people that are on the line,” Joyner said. “AI is not going to be accountable. People are accountable.”

Madeline and Elvin Eaton operate on a much smaller scale than Joyner and Westerbeek, but they shared similar sentiments about the importance of having people involved in handling customer relations and other aspects of their operation.

The Eatons started their 12-acre Fairport Farms in Granville County, N.C., in 2022 as a retirement venture. They focus on growing and marketing nutrient-dense produce in one of the state’s most food-insecure regions.

The two now use AI to help with pest identification, advertising and interacting with their community-supported agriculture customers, and they look forward to the day when they can use AI to get real-time, actionable data on soil management and water efficiency.

“The community-based connection is something that AI can help you with,” Madeline Eaton said, “but must ultimately be in the hands of the farmer.”

‘It Has To Be Economical’

While the Eatons shared optimism about the future of AI on the farm, Elvin expressed concerns about the price tag.

“Within the next five years, we would love to use AI to its fullest,” he said. “At the same time, we do have to remember that some of the technology that’s … available now — we really can’t afford it. … It has to be economical.”

The disjuncture between promise and price came up again when a panel of agricultural technology investors took the stage. Frank Klemens, managing director for the Generation Food Rural Partners Fund at Big Idea Ventures, said the technology that has taken hold on farms so far isn’t the flashy stuff — it’s the AI that handles back-office tasks.

“Farmers want that,” Klemens said of AI tools that help with inventory, invoices and pay schedules. “They are torn in so many directions. … They don’t want the AI to control their money. They want their AI to make their lives simpler.”

Ruggedness and Reliability Matter

Alongside affordability and simplicity, speakers insisted that agricultural AI must be rugged enough to survive not just in a controlled lab but on an actual farm.

One of the conference’s keynote speakers, Sebastián Villegas, of the livestock tech company Asimetrix, said the company had to specially engineer cameras for farms.

“For example, (with) Pig Vision … we process most of the images inside the camera, because … processing it in the cloud is very expensive. You need high-speed internet,” he noted. Also, “electrical outages out in the field are very common. Lightning might strike the farm, so we need to protect it with an internal battery so they don’t get burned.”

Reliability, not just ruggedness, is also critical.

“Technology has to work all the time, and some of our technologies aren’t there yet,” said Angela Green-Miller, a University of Illinois Urbana-Champaign animal behavior researcher who was part of the Precision Livestock Panel.

“They work part of the time, and then there’re places where they fall apart. And producers need it to work all the time,” she said.

Early failures, she added, have left the industry wary, not more receptive: Producers “essentially traded one headache for another,” she said, and “need to know it works before they kind of jump in the investment space again.”

Defining ‘Responsible’: Privacy, Foresight and No Silos

While it was clear that economic considerations are the bottom line when it comes to farmer adoption of AI tools, what a truly responsible AI system in agriculture would look like was more nebulous. However, three ideas surfaced repeatedly: Farmers’ data must stay in farmers’ control; the tools need to move from reacting to problems toward anticipating them; and the industry’s sprawling platforms and equipment must start talking to each other.

Wariness around data privacy came up during discussions of precision livestock production. Villegas, of Asimetrix, noted that hog and poultry producers compete fiercely with one another, and a leak of the wrong data could be devastating.

“Producers, they are very competitive among each other,” he said. “They don’t want the numbers to be out there.”

During the Digital Agricultural Institutes panel, Renata Ivanek argued the fix isn’t more restrictions but better architecture. Ivanek advocated for federated learning, which helps with data privacy by keeping raw data on local devices, sharing only model updates and combining these updates on a central server. Instead of moving raw data to a central cloud, the AI model is sent to the local data.

“I think what we need as a solution is a way to integrate data without actually sharing data,” said the co-director of Cornell Institute for Digital Agriculture.

Beyond Chatbots to Predictive Tools That Put Farmers Ahead

Beyond data privacy, conference participants also voiced a need to move from reactive technology to proactive systems.

Syngenta’s Zach Marston, who leads development of the company’s Cropwise AI platform, argued that today’s chatbot-style tools are just waystations.

Marston argued that farmers need systems that automatically alert them to issues, rather than just answering questions — ones that might flag, for example, that “based on last year’s yield data and spring soil moisture, three fields need a different hybrid.”

Because a farmer’s planting decisions can be locked in well before the growing season starts, the industry needs to “understand what a producer needs early enough to deliver it before application,” he said.

During the Precision Livestock Panel, Isabella Condotta, a University of Illinois Urbana-Champaign assistant professor of animal sciences, argued that traditional farming relies on a “reactive approach,” where managers only intervene after seeing visible symptoms of illness.

She sees AI as a way to move the industry toward a “more proactive approach” to understanding normal animal behavior and physiology, providing continuous monitoring and detecting deviations from the norm so that farmers can intervene before symptoms are visible.

The Importance of Having Applications and Systems That Communicate With Each Other

The third ingredient for responsible AI in agriculture — interoperability — was perhaps voiced more insistently than any other single idea at the conference.

Gyami Shrestha, an Earth system scientist and advisor who previously served as a U.S. Department of Agriculture national program leader and director of the interagency U.S. Carbon Cycle Science Program, framed it as a central challenge that presents opportunities for cross-disciplinary collaborations.

“We are not building systems yet. … We are building intelligent silos,” she said, warning that AI in agriculture “is not a tech challenge, it’s a systems challenge.”

Researchers and extension educators recounted incidents where unaligned technologies were causing frustration among farmers.

As Condotta, of the University of Illinois, noted, “If you have multiple different technologies that do great things for your farm, but at the end of the day they don’t talk with each other — that’s a big complaint that we get. If it’s going to just become noise in a farmer’s life, then they just shut down.”

Mohan Tavorath offered similar thoughts during a panel session with ag tech investors. Tavorath noted that for AI to be useful, it must be integrated; farmers cannot be expected to consult five different platforms from five different companies to make one decision.

“A lot of startups … come and say, ‘Look at everything we can do. It’s only a dollar an acre.’ But if there’s 10 of them, that’s $10 per acre, and nobody’s integrating those solutions together,” he explained.

Getting There: Trust, Testing and Talent

While a picture emerged of what responsible AI in agriculture should look like, converging around three prerequisites: earning trust deliberately, testing for reality rather than the lab and building a workforce fluent in both agriculture and AI.

On trust, Tavorath mentioned crop consultants who have quietly earned decades of credibility that no computer algorithm will inherit.

“That’s trust that has been built up over 35, 40 years,” he said. “They’re aging out, and that’s something that nobody talks about. That trust doesn’t get built immediately.”

The opportunity for AI, several panelists contended, is to capture and extend that expertise rather than replace it — an idea Westerbeek raised: “I think a big opportunity is to gather that knowledge from those folks and be able to transfer it down to the newer folks in the industry at a much more rapid pace than maybe we’ve been able to do in the past,” he said.

Moving Beyond the Lab and Into the Field

When it comes to the importance of testing for reality, Zach Marston offered hard-won advice.

After Cropwise AI’s early herbicide-recommendation feature scored just 66% accuracy — made worse by the fact that the system would, as he put it, “confidently cite correct documents while reaching incorrect conclusions” — his team built what he called a field-readiness framework.

The framework tested for a range of factors, including accuracy, repeatability and robustness in the hands of real users.

“Test for reality, not the lab,” he encouraged the innovators at the conference. “The feedback that stings the most — that’s where your signal is.”

During a joint plenary session, Chris Reberg-Horton, an NC State crop science professor, and U.S. Department of Agriculture’s Steven Mirsky described a parallel effort in computer vision. Their massive, open-source Ag Image Repository was built to provide data that all players, including university researchers and small companies without deep pockets, can use to develop AI-enabled tools for agriculture.

The repository is available to any U.S. researcher through the USDA’s Scientific Computing Network, or SCINet, and the plan is to make it even more widely available.

“We really have to step in and govern our own destiny for how we do this,” Reberg-Horton said. “If we just wait and let the private sector tell us how we collaborate, then we might not like how that’s done.”

The Need for ‘Bilingual’ Graduates Who Can Deliver Results

Conference speakers also focused on the importance of focused workforce development. Nearly every university panelist at the conference described the scramble to train graduates fluent in agriculture and AI and capable of translating research into practical solutions.

They highlighted a range of initiatives across their institutions, from K-12 toolkits and undergraduate minors to online graduate programs and student scholarships. In addition, they emphasized the need for collaboration with industry and across disciplines to address complex challenges.

Tami Brown-Bandl, professor of biological systems engineering at the University of Nebraska-Lincoln, described the need for a future workforce training in computer science, animal science and social science to advance precision livestock production.

Ignacio Ciampitti, the moderator of the Digital Ag Institutes Panel, was among them. The Purdue University professor of agronomy emphasized the need for universities to start talking to each other and to industry about training graduate students for the AI-in-ag workforce.

“When is the moment that we’re going to start talking to each other to coordinate, to make sure that when we are doing training of grad students, we are being intentional, we talk to industry, we integrate industry from the beginning to say, ‘What is the type of skills and workforce that you need to make sure that we are ready to train that type of workforce?’” he asked.

Others on the panel pointed to a critical need for better entrepreneurial training and support ecosystems for both students and faculty members.

Ajay Sharda, co-director of Kansas State University’s Institute for Digital Ag Advanced Analytics, noted that while faculty are expected to perform high-level research and secure funding, the academic ecosystem does not train them to be entrepreneurs. Thus, researchers sometimes come up with promising solutions that don’t meet industry needs, and other times, they lack the knowledge to protect their intellectual property effectively.

“We don’t train people to be entrepreneurs,” Sharda said. “I think those are the things we need to facilitate.”

NC State Prioritizes AI in Ag

If the conference produced a consensus about what responsible AI in agriculture should look like and laid a foundation for how to get there, its host institution, NC State, offered a working case study for accelerating AI applications in agriculture.

Garey Fox, the dean of NC State’s College of Agriculture and Life Sciences (CALS), said there is a push at NC State “to scale AI and ag tech as quickly as we possibly can.”

“We are believers … that AI is going to help lead us into that next generation of agriculture,” he added.

He also outlined a strategic framework for the future of AI in agriculture that he called “science-integrated AI.” The approach aims to better understand the highly complex biological mechanisms underlying agriculture by using the principles of physics, chemistry and biology instead of just guessing patterns based on past data.

Fox highlighted significant investments that NC State has made in faculty positions, facilities and infrastructure to support science-integrated AI.

The new Genome Editing Center for Sustainable Agriculture is using “AI to help with the precision, the speed and the efficiency of our CRISPR technology,” he added. “You’re going to see significant movement … in our college in terms of the rate which we hopefully can release new varieties.”

AI, paired with other ag tech, is also being used to address issues in food animal production, which constitutes the largest sector of the state’s $117 billion agriculture and agribusiness industry.

And the N.C. Plant Sciences Initiative (N.C. PSI), operating from NC State’s Plant Sciences Building, is “doing amazing things … with the help of a focus on AI,” he said.

Closing the Lab-to-Farm Gap

Following the conference, N.C. PSI Executive Director Adrian Percy noted that the initiative is well-positioned to help close the gap between promising research and AI-enabled tools that farmers can trust.

“Universities are strong at proof-of-concept research,” Percy said, “but existing technology pipelines are not designed to move AI tools into real-world, maintainable applications for farmers. The university must step in to build last-mile infrastructure for agriculture.”

The N.C. PSI, a joint effort of CALS and the university’s Office of Research and Innovation, is designed to go that last mile, Percy added.

He pointed to AI-powered tools like the Agriculture Image Repository, BeanPACK and Sweet-APPS as proof of the N.C. PSI’s ability to deliver. BeanPACK is a free, web-based decision-support tool that helps soybean producers decide on ideal planting dates and maturity groups, and Sweet-APPS uses advanced imaging and machine learning to help evaluate sweetpotatoes by size, shape and surface quality.

In fact, Thomas Joyner mentioned that Nash Produce is working with the Sweet-APPS team to develop AI-enabled tools to analyze specific bins of sweet potatoes to determine their best market fit. This allows the packing operation to distinguish between produce that meets the high aesthetic standards of retail outlets and produce better suited for food service providers, helping ensure that the packs gets what they can out of every bin, reducing waste and increasing profits for growers and the company.

The emerging REFRAME project — Resource Engineering Framework for Responsible, Augmented Modeling and Engagement — also aims to help close the gap, breaking down the kind of “intelligent silos” Shrestha pointed out.

The goal is to better inform supply chains so they can create opportunities from agricultural leftovers. The project draws on the expertise of Jones, the N.C. PSI’s education and workforce development platform director, and other N.C. PSI faculty affiliates in disciplines ranging from electrical and computer engineering to plant and microbial biology.

The open-source platform will connect fragmented supply chain models to help farmers, scientists and businesses evaluate and convert farm residues, such as misshapen sweetpotatoes and their green tops, into profitable products.

The long-term payoff of more widespread AI adoption in agriculture will be revolutionary, Percy concluded.

“AI-enabled tools will allow farmers to see earlier, respond faster and manage more precisely than ever before,” Percy said. “This is not incremental improvement. It is transformational.”

Jones agreed, noting that ongoing conversations like those sparked at the conference will help speed innovation.

“Bringing these groups together — our universities, extension programs, startups and partners from industry and government — accelerates learning and supports adoption of practical tools that can shape the future of agriculture through AI,” she said.

Judging by the discussions at the AI in Agriculture conference, how that future unfolds will come down to whether AI-enabled tools can pass the same tests the Growers Perspectives panelists laid out: Does it stand up to harsh on-farm conditions? Does it pay for itself? And does it leave those who know the land best in charge?

The conference included a technical poster session and technical talks on over 250 AI-related projects. Conference participants also had the opportunity to tour Syngenta in Research Triangle Park and NC State’s Lake Wheeler Road Field Laboratory and Plant Sciences Building in Raleigh. Four preconference workshops were offered: From AI Pest Detection to Scalable Deployment: Hands On YOLO and AWS for Agriculture; Precision Sustainable Agriculture Tools: Production-Ready Decision Support; Agricultural Analytics with SAS Viya: Scalable Data and Modeling and AI‑Ready Pesticide Labels: Building the Digital Foundation for Smart Agriculture.

Acknowledgments

The 2026 AI in Agriculture Conference focused on advancing artificial intelligence and data-driven innovations for resilient and competitive agricultural systems. Diamond Sponsors: John Deere, Syngenta, and Schmidt Sciences Foundation. Platinum Sponsors: AI-LEAF, NC Biotech Center, Bayer, Prairie View A&M University. Gold Sponsors: Cotton Incorporated, InstaDeep, AMADAS Industries, Purdue University, Texas A&M University, and Mississippi State University. Silver Sponsors: Asimetrix, Hiphen, Envu, Center for Digital Agriculture at University of Illinois Urbana-Champaign. In-Kind Sponsors: N.C. Plant Sciences Initiative, Data Science and AI Academy and the Center of Excellence for Regulatory Science in Agriculture at North Carolina State University.

Bayer, the NC Biotechnology Center and the USDA National Institute of Food and Agriculture supported the workshops within the conference. NIFA’s support came through the Food and Agriculture Cyberinformatics and Tools program for the DSFAS Workshop Series, From Integration to Innovation: Advancing Agriculture through AI at the Nexus of Science (Award No. 2026-67021-46565; Accession No. 1034690) and the Learning to Leading: Cultivating the Next Generation of Diverse Food and Agriculture Professionals (NextGen) Program, From System Approach to Promoting Learning and Innovation for the Next GenerationS in Food, Agriculture, Natural Resources and Human Services (Award No. 2023-70440-40153).

© 2026 Alliance Partner | Agricultural Innovation & AI Solutions