Will Autonomous Harvest Reach the Goal Line?

Some in the industry liken agtech automation to a series of football plays, running the ball down the field until funding runs out — as full automation has remained elusive.

By Christina Herrick

•

October 26, 2025 07:00 PM

Editor’s note: This is the latest story in a series exploring the current state of labor in the fresh produce industry. This is the first of two pieces exploring the role of technology and its relationship to ag labor.

Technology’s role in the future of specialty crop farming isn’t new. In fact, Grady Avuil of Auvil Fruit and Tom Mathison, the founder of Stemilt Growers, founded the Washington State Tree Fruit Research Commission in 1969 in part to research the future of autonomous harvesting.

Though significant progress has been made on both autonomous weeding and other tasks, harvesting continues to remain elusive. While labor costs remain high and labor availability is still a significant challenge, agriculture technology companies have looked to all parts of the growing process — from record keeping, to spraying and more — to reduce the need for labor.

Some liken the process of harvest automation development to a series of football plays: Each tech startup or company runs the ball further down the field only to eventually run out of funds, then the next tech company picks up the ball and runs it further down the field and eventually runs out of funding, and the process goes on and on.

But the real question is, will harvesting automation get to the end zone?

Harvest Challenges

For starters, it’s one of the biggest challenges to solve in the agtech space, says Walt Duflock, vice president of innovation at Western Growers.

“Harvest is the most capital-intensive of the activities,” he says. “Crop destruct with weeding and thinning is pretty simple. I pick which ones I want and the ones I want to destroy. With harvest, it’s very skilled labor. Your ability to identify the right plant, fruit or vegetable, pick it, put it in a bin and get that bin to the truck; there’s a lot going on there.”

Daniel Butler, regional sales manager of RDO Equipment Co., says harvest automation is “the lowest hanging fruit,” but while it’s a significant chunk of labor costs, some crops are more prime for automation than others.

“There are a few commodities that are easier to mechanically harvest, but lettuce remains one that’s challenging,” he says. “It’s super easy to damage.”

Like lettuce, so many crops are delicate and labor-intensive to harvest, which adds to the challenges of automation, Butler says. While green onions, snow peas and snap peas are quite labor-intensive, cabbage and radishes are much easier to mechanically harvest.

“Cabbage is much more robust,” he says. “You can peel a few layers off the outside that have become damaged and still have a nice core head of cabbage. We are seeing some romaine being harvested mechanically now. One of the biggest challenges with the romaine is the cut point is right against the soil. And so soil contamination was a big part of holding that back.”

Butler says something as nuanced as strawberry harvest is incredibly difficult to mechanize and automate.

“A strawberry is just so delicate, and a marketable strawberry is picked with the green crown on,” he says. “Is it the right color? Is it the right size? There’s a small twisting motion that goes into the picking.”

The other struggle, too, is that these harvest innovations not only have to meet the accuracy and efficiency of hand harvest, but they also have to exceed hand-harvest for it to make economic sense for growers.

Sean Sundberg, business integration manager for high-value crops at John Deere, says as labor costs and the cost of growing continue to rise, so too does the need for innovative solutions.

“I can’t think of another segment within the agricultural community that is not necessarily starving for some of this, but they need this,” Sunberg says. “This piece of the market, in many ways, has not experienced the technological evolution that’s occurred elsewhere.”

Economic Drivers

Duflock says many agtech companies entering the market to create an autonomous harvest solution don’t understand the significant capital it takes to get the idea to the finish line.

“Not everyone who starts a harvest robot realizes it’s going to be a $50 [million] to $100 million endeavor,” he says. “They’ll raise a little bit, they’ll do some work, they’ll raise a little more, but eventually you get to that spot where, if my grower is not acknowledging that I’m on the right path to get it to an economic situation that works, I’m in trouble.”

That’s the unfortunate story many agtech startups face, he says — the disconnect between innovation and the economic requirements for practical application. Duflock says this goes beyond just believing in an agtech solution; growers need these solutions to help with profitability. He says he’s had growers tell him, “I can’t create dollars for a startup. They have to help me pull dollars out so I can spend on them.”

“People don’t realize the growers have so much input, because they are the source of truth,” he says.

As nuanced as production is for each crop, that means that an autonomous solution would need to be just as targeted, Duflock says.

“You’re doing a lettuce harvester; you’ve got one robot,” he says. “You’ve got an apple harvester; it’s a different robot.”

And because these startups need to create an individualized solution for each crop, that makes it hard for them to make the math work to stay in business.

“The economics of the startup are challenged because you can only service the one crop per build,” he says. “Economics of the grower are harder because they can only work with one crop per robot. And the economics for the capital, for the venture capital, is providing the money, is challenged as well, because, again, if they’re going to give you $3 [million], $5 [million], $10 million over the course of a financing round, they really want to work with more than one crop, but they can’t for the same reason.”

Duflock says the agtech industry in the U.S. has seen a significant shift in venture capital investment from harvest and harvest-assist startups to more investment in nonharvest automation.

“Part of the reason what’s happening with harvest is happening, even though it’s two-thirds of the labor hours, is when you see success in the weeding robots, the thinning robots, the planting robots, the harvest-assist venture capitalists will go toward where there is some success,” he says. “There is no success right now in fresh harvest. There is success in fresh nonharvest.”

Realistic Expectations

Duflock says Western Growers and California Polytechnic State University looked at the economics of specialty crop production in the state.

“California is spending $16.3 billion on 850 million hours of ag labor a year,” he says. “And depending on the crop type, about two-thirds of those hours are harvest and one-third of those are cultivation and other activities.”

And currently, he says, growers only automate about 3% of those nonharvest tasks, but he says adoption could reach 15% to 20% soon.

Butler says while agtech has been at the forefront of dealers’ minds for 15 years or more, grower adoption has been more sluggish. He says that is interesting, because in contrast, processing and packing facilities have become highly automated and mechanized.

Butler says, truly, the first foray for many lettuce growers into automation was the introduction of thinning technology.

“A few technologies came out that would chemically thin the crop using usually a hot shot of fertilizers, so they don’t damage the keeper plants and that made a huge impact on what labor we could put in,” he says.

Thinning technology became a launching pad for growers to look at automated weeding and more.

“What else can we do mechanically in the field to get people out of the field?” he asks.

Sundberg also says John Deere’s partnership in The Reserovir, an on-farm incubator platform, hopes to tackle some of these barriers to agtech adoption on a wider scale.

“If we can find a way that’s maybe less capital intensive, maybe less labor involved, maybe less risk for them,” he says. “That’s kind of what we’re all trying to accomplish.”

And as for automating harvest?

“We’ll get there,” Sundberg says. “It just tells me that that’s where the hardest problems to solve are.”

Your next read in this series:

Labor