For the last three years, I’ve spent much of my personal and professional time trying to understand what GenAI means for agriculture. And I’ve reached a point where I think someone should point out what most of the hype conversations are missing.

Don’t get me wrong—I still believe GenAI has the potential to become the technology that fits naturally into the way agriculture already communicates and operates. Notice that I said GenAI, not AI as a whole. That distinction matters.

One of the biggest mistakes I see today is treating AI as a catch-all solution for every problem. While there have been meaningful advancements across the broader AI landscape, the breakthrough that captured everyone’s attention came from large language models built on transformer architectures, culminating in experiences like ChatGPT 3.5. If we’re talking about AI, we need to be specific about what we mean.

What Makes GenAI Different for Agriculture?

GenAI is unique because transformer models aren’t limited to language. The same foundational technology can generate text, images, video, and voice. That versatility creates opportunities that align surprisingly well with how agriculture already works.

Here’s the part that Silicon Valley often misses:

In an industry where dirty hands, windshield time, and bright sunshine on screens are the norm, forcing people to spend more time interacting with software often feels backward. Pen and paper are frequently more convenient than many digital tools.

This is where GenAI becomes practical. Instead of forcing users to adapt to software, it enables software to adapt to the way people naturally communicate.

Imagine a future where equipment simply understands what’s happening. A sprayer recognizes a product label through onboard cameras. As soon as the operator gets back in the cab, the system asks for confirmation rather than requiring manual data entry. The technology needed to make this happen isn’t nearly as far away as many people think.

GenAI’s greatest promise may be its ability to fit into existing agricultural workflows rather than requiring entirely new ones.

The Wrong Question

Most of the GenAI coverage I see focuses on how intelligent the models have become.

Newer models can answer a remarkable percentage of Certified Crop Adviser (CCA) exam questions correctly. That’s impressive, but my question is simple:

Why does that matter?

Where exactly will we use that capability?

The answer isn’t found in the technology itself. It’s found in understanding how decisions are actually made in agriculture.

Trust in farming is built over years through relationships with experts—not platforms. I’ve heard stories from leading precision agriculture companies that developed recommendation engines which outperformed agronomists in field trials, yet growers still preferred advice from trusted people.

That’s not a flaw in the technology. It’s a reality of our industry, and it’s likely to remain true for quite some time.

Just Because We Can Doesn’t Mean We Should

This leads to a fundamental question:

Just because we can automate something, does that mean we should?

Simon Sinek’s Golden Circle applies here. Start with why you want to automate something—not simply what can be automated.

When evaluating opportunities, I like to place jobs-to-be-done into three categories:

• Tasks you wish would simply disappear (repetitive and mundane work)
• Activities you’d focus on if you suddenly gained three extra hours every day
• The unique work your top 10% performers do better than everyone else (“hero work”)

If GenAI can help eliminate the first category and create more time for the second, you’re creating real value for users.

I wouldn’t recommend starting with the third category. That may be the long-term vision, but it’s rarely where the immediate opportunity exists.

The Unsexy Opportunities Are the Most Valuable

The biggest wins aren’t flashy. They’re operational.

Think about what your best salesperson would do with three additional hours every day. They probably wouldn’t spend it entering notes into a CRM. They’d spend it building relationships and helping customers.

So why are so many companies focused on automating relationship-building instead of automating the administrative work surrounding it?

Imagine a CRM assistant that allows a salesperson to call a virtual agent while driving between appointments and verbally update customer records. Or a system that instantly identifies customers whose prepay activity is less than 40% of the previous crop year.

Those aren’t headline-grabbing innovations. They’re the types of solutions that actually move the needle.

Why Human-in-the-Loop Matters

Large language models will continue improving, but they will always face challenges around accuracy and hallucinations. The issue is often overstated, but it’s real.

That’s why human oversight remains critical.

In agriculture, this is actually an advantage. Relationships are still built through people, and human judgment remains an essential part of decision-making. Rather than viewing human involvement as a limitation, we should see it as a feature.

A practical design approach is simple:

• Let the AI make its best recommendation
• Clearly show the user what it did and why
• Allow the user to easily adjust or correct the output

Good systems don’t hide the AI’s work. They make it transparent.

The Psychology Problem Nobody Talks About

There is another challenge that many companies overlook: psychology.

Ask yourself a simple question:

Have you ever truly enjoyed interacting with a chatbot?

For most people, the answer is no.

When a GenAI product feels like a traditional chatbot, users immediately associate it with every frustrating chatbot experience they’ve had before. Robert Cialdini discusses similar concepts in Pre-Suasion. These mental anchors matter.

The challenge becomes even greater when users are familiar with ChatGPT. They begin expecting every GenAI experience to have the same capabilities and flexibility.

Whether those expectations are realistic doesn’t matter. The expectation itself shapes perception.

Start With the End in Mind

A comment from Olivier Harvey has always stuck with me:

“Most people aren’t creative.”

That’s important because today’s LLM experiences often start with a blank page. Users are given unlimited possibilities but very little direction.

For many people, that’s overwhelming.

Another quote that has influenced my thinking comes from Joe Middione:

“Start with the end in mind.”

Figure out what outcome your users are trying to achieve. Design the experience around that outcome. Then let the LLM handle the complexity behind the scenes.

Users shouldn’t have to figure out how to use AI. The product should guide them toward success.

The Companies That Win Will Understand This

The companies that win in agriculture won’t necessarily have the smartest models.

They’ll be the companies that understand GenAI’s greatest value isn’t replacing human judgment.

It’s giving people their time back.

Time to build trust. Time to make nuanced decisions. Time to focus on the work that only humans can do.

And in agriculture, that may be the most valuable outcome of all.