[This article was written by Dee Shore and originally published by CALS News.]

Around a dozen North Carolina Cooperative Extension agents came together recently to explore ways to digitize their century-old mission, learning about AI tools that could enhance their delivery of research-based knowledge to farmers and families statewide.

As part of the 2026 AI in Agriculture Conference held March 31 to April 2 at NC State University, the agents took part in a standing-room-only Hands-On AI for Extension: Using Generative AI With Agricultural Data workshop.

Supported by NC State Extension and the N.C. Plant Sciences Initiative, the workshop was the second AI-related Extension agent workshop offered by SAPLINGS, a multi-institutional project led by N.C. A&T State University and funded by a U.S. Department of Agriculture initiative.

NC State is a major partner in the project, providing students and Extension agents with opportunities to learn more about AI and data science. These SAPLINGS efforts are part of the ag data science training, education and outreach led by N.C. PSI’s Data-Driven Plant Sciences Platform.

NC State Libraries’ Shannon Ricci and Kristy Borda led the April 1 workshop, demystifying terms related to AI, outlining AI’s powerful capabilities and significant pitfalls, and demonstrating ways it could be put to use for Extension education.

Generative AI: A Tool for Efficiency

In North Carolina, Extension agents serve as the bridge that connects people to research-based knowledge and technology from NC State and N.C. A&T universities.

Agents operate out of 101 local Extension centers throughout the state, providing timely educational opportunities tailored to local needs, with many demands on their time.

One day, they might be delivering hands-on workshops and preparing educational articles, newsletters and videos, and the next they might be making farm visits to offer one-on-one technical assistance.

In between, they’re often answering and returning phone calls and emails from individuals who need trustworthy advice to sort fact from fiction in a world increasingly saturated with information.

N.C. PSI Extension Outreach and Engagement Platform Director David Suchoff, a faculty member in NC State’s Department of Crop and Soil Sciences, saw the one-and-a-half-hour workshop as a step toward helping agents realize the promise of generative AI to help them in their work.

Generative AI models learn patterns from existing data to create original content, and the workshop was designed to allow agents to learn about tools that could reduce the time agents spend on tedious tasks, allowing them to focus more on what Suchoff called “the human aspect of their job — interactions with their stakeholders, building trust, problem solving.” 

Different Tools for Different Tasks

Borda and Ricci demonstrated two AI tools from Google that could help agents achieve such efficiency.

Borda, research librarian for sciences and instructional strategy, demonstrated how Gemini could be used to cut down on the time it takes to reformat or clean up basic text or data and brainstorm and draft documents such as short quizzes for field day participants and scientific research summaries for general audiences.

Ricci, the Libraries’ associate head of data science services, focused on NotebookLM. Unlike general purpose AI models, such as Gemini, that synthesize information from massive datasets, NotebookLM’s responses come only from documents the user provides, minimizing errors commonly referred to as hallucinations.

After uploading large agricultural production manuals, Ricci demonstrated how the tool could accurately answer specific technical questions, create tables and generate podcasts, videos and other content, while providing precise citations pointing to the source documents.

The workshop leaders also covered NC State policies regarding AI use and data security and the importance of respecting copyright and intellectual property and being transparent about their use of AI.

The Need to Retain the Agent’s Human Touch

As the workshop unfolded, agents offered ideas for how they might use AI to make their work more efficient – from drafting and finetuning funding proposals to creating professional maps from hand-drawn ones.

They also cited ways they’d already used AI tools. One participant described using AI to create a slide deck for a mid-year financial review, while another had used AI to scan and transcribe nearly unreadable cursive from old family genealogy documents.

Borda and Ricci encouraged the agents to continue exploring and testing ways to use university-provided AI tools in their work. They also emphasized the need for agents to incorporate a human touch and their subject-matter expertise when using AI.

As Borda put it, generative AI is simply a tool, but not a solution.

Generative AI won’t “give you the full picture,” she cautioned. “It’s not going to know what context it’s leaving out. It will not tell you when your idea is bad … It’s a cheerleader. It will not tell you when it doesn’t know the answer to your question or when it lacks the training data to answer your question.

“And it will not produce original ideas. That is part of …  our value as humans,” she said. “We are the innovators.”