Robots are Coming to Harvest your Food
by Twilight Greenaway  /  September 29, 2020

“It’s a misty morning near Salinas, California and the TX harvester—a lightweight, driverless tractor covered in canvas—is picking strawberries. Like a dot-matrix printer moving along a page, the harvester’s robotic hands move back and forth along the beds, scanning for signs of red. When it identifies a ripe berry, it dives down, gently plucks it from the plant, and places it in a crate. At the same time, a crew of about two dozen farmworkers is also harvesting strawberries just a few hundred feet away, on an adjacent farm. As an energetic song blasts from a parked vehicle, the men and women stoop to pick berries straight into plastic clamshells that they tile, side-by-side, into cardboard trays. Once their trays are full, the workers take them back to one of several sorting tables spread out along the access road (to allow for social distancing). It’s clear by the speed at which they’re walking—and in some cases even running—to drop off each box that these men and women are getting paid by the piece.

I’ve trudged through the muddy, irrigated fields to watch both forms of harvest with Kyle Cobb,’s youthful, clean-cut CEO. The company was the first to mechanically harvest strawberries for commercial sales last year, and had raised just under $10 million by June 2020, including a $7.5 million Series A round in 2019. After building a robot that cleaned solar panels, Cobb and his team dove into agriculture, where they hope to put an end to the notoriously grueling, repetitive work of harvesting strawberries. If things go as planned, and is able to scale up over the next several years, Cobb says, “You’d see the same crew, but instead of it being this big, you’d see about half the size . . . and they wouldn’t be doing the traditional picking like this. They’d be doing a combination of sorting and packing in a very comfortable ergonomic set up.” Today, instead of the fleet of three or four harvesters that are typically picking berries, the TX is in the field prototyping, gathering data to be used by the company’s team of engineers at their office three hours north, in Davis. Prototyping is slow, exacting work, and the machine is accompanied by field operations manager Jorge Cava, who carries a tablet and watches patiently as the harvester moves along the rows, learning thousands of iterations of berry, stem, and leaf. “We should get several hundred more hours testing on it,” said Cobb, before it goes back into the field.

Compared to the hustle taking place on the next farm over, it’s a pretty low-key scene—boring even, to the untrained eye. And yet, Cobb, Cava, and others working to automate the harvest have been in the midst of their own hustle over the last few years. Now, the pandemic has ratcheted up the pressure. For farmers considering investing in the automation, Cobb tells me in the field, he sees the pandemic as one of several factors that will breaks the camel’s back. “It’s the rising cost [of labor], it’s the already-dwindling supply, the aging of the workforce, the hard work. Add in a health pandemic that further limits the supply and complicates your daily logistics, and automated harvest starts to sound really nice,” he said. Hazardous conditions caused by this year’s wildfires may also be a factor, although they haven’t stopped many crews from harvesting this fall. It’s not just growers who may soon embrace the technology. In California, most of the counties with the highest rates of infections are in the Central Valley, the state’s most productive agricultural region, and home to hundreds of thousands of farmworkers. And as farmworker communities around the country battle a growing number of coronavirus outbreaks, illnesses, and deaths, the discussion of automation across the food production spectrum has grown in the public arena as well. If the people doing the work on farms are getting sick, the logic goes, why not just replace them with machines?

The transition for the companies isn’t going to be fast or easy: Cobb estimates that it will likely take five to 10 years before it’s really complete. “We’re working as hard as we can,” he said, as he details the many challenges companies like his face in the process to get the machines out in the field, working as fast as human pickers. For a good part of the summer, for instance, the strawberry plant’s leaves grow so large that they essentially block the harvester’s vision from above. And strawberry breeders have so many other priorities, that it could be a while before they start breeding plants that make it easier for the harvesters to do their job. Of course if—or one of the other companies in its lane—succeeds, the shift won’t be easy for farmworkers either, nor for farm-centric communities such as Salinas, Watsonville, and Oxnard. The strawberry industry employs about 55,000 people in on-farm jobs on an estimated 38,000 acres in California—making it one of the state’s more labor-intensive crops. And if automation successfully cuts that number in half it could mean the loss of over 27,000 jobs in that slice of the produce sector alone.

For Sébastien Boyer, the shift toward increased automation in farming is an inevitable one. Boyer is the CEO of Farmwise, a company that launched its first autonomous weeding robots in early 2019, and has grown quickly in the year and a half since then. FarmWise went from having a small handful of weeding machines in 2019 to 20 of them in 2020. It also scaled out from a startup in a garage in San Francisco to a 700,000-square-foot shop and headquarters in Salinas. And Boyer says he has seen an increase in interest from farms in California and Arizona. “We see a kind of short-term positive shock in the attractiveness of what we do. But we are also seeing increased discussion around automation,” Boyer told me in his thick French accent. “By and large, what I think is going to happen during the crisis is a faster push for things that makes the overall supply chain less reliant on the uncertainty of manual work being done in the fields.”’s Cobb echoed this sentiment. “There are always early adopters, and they have been ahead of this trend for all kinds of reasons,” he said. “But now that second wave of people, the mass adoption . . . I think they’re moving faster. They’ve changed their mindset faster than you otherwise would have expected.” Emily Reisman is an assistant professor in the Department of Environment and Sustainability at the University of Buffalo and a recent transplant from the University of California, Santa Cruz. As part of a larger effort to document and study the agtech industry with a group of other researchers, including U.C. Santa Cruz’s Julie Guthman, she has been attending agtech events—which have moved on online but not slowed down—since before the pandemic began. “I think it’s unlikely that COVID will dramatically accelerate the development timelines of these companies, especially for mechanical devices,” said Reisman. “But this moment might allow certain technologies to gain legitimacy and potentially additional financial backing, institutional support, or broader public acceptance.”

Reisman is also concerned that COVID is being used as a way to depoliticize technology that displaces workers. She has spent the last few years studying the almond industry, where automation already allows for near-instant harvest thanks to mechanical tree shakers that can remove an entire tree full of nuts in around a minute. On those farms, the number of people needed per acre is minuscule. “I found that the crop’s high level of automation is part of what makes it so attractive to financiers looking at land as an investment. Low labor means low risk,” she said. “So, I think automation is attractive not only to farmers or technology companies, but also people who are interested in land as a financial asset.” Rather than selling their equipment, both Farmwise and contract with farmers to pay for the machines’ services—which allows the companies to send fleets of weeders and harvesters around the area. “I found that [almonds’] high level of automation is part of what makes it so attractive to financiers looking at land as an investment. Low labor means low risk.” And with new overtime laws for farmworkers going into effect in 2022, Cobb says the investment in automation is “more of a hedge for future cost inflation rather than a significant cost reduction.” Indeed, automated harvesting will potentially do away with the limitations of the workday. It’s not typically safe to employ people to work on farms at night—but machines like the TX harvester don’t care about light or temperature, nor do they have circadian rhythms; they can conceivably run for 24 hours if needed. “It’s filling the gap in two ways. One is just by supplying machines that can pick instead of humans, and two, improving quality of work for the humans who are left so that more people are attracted to the line of work than are today,” said Cobb, who envisions a transformed industry unburdened by the kinds of repetitive, body-ruining work that is so common in today’s fields.

When I spoke with Farmwise’s Boyer in April, right after the coronavirus hit, he told me his company was in a rare position to be hiring several people as they ramped up their customer base. “We’re paying significantly more than the average wage that fieldworkers make today. And that’s because we’re going to make every one of those workers drastically more productive than if we were asking them to do this work manually,” he said. Jaime Eltit, Farmwise’s commercial operations manager, says the new, better-paying jobs created by companies like his are an important response to the farm world’s “shrinking and deteriorating labor force. Probably the youngest people that you see out there right now are around my age,” said Eltit, who is in his 30s. “But the generation below them, those kids aren’t going out into the fields. This kind of work is hard; it’s not really desirable. And so there are going to be less people, but [a small number of] more skilled people doing the job of others.” And companies like Farmwise are “replacing the jobs at the bottom,” such as thinning lettuce, weeding, and harvesting, he adds. In fact, nearly everyone I spoke to in the agtech industry preferred to focus on the “better jobs” aspect of the coming shift. When I asked Cobb about the fact that the future he and others envision could involve fewer jobs, he cautioned me to be careful about that phrase. “Right now, one of the ways [farms] bring people to bridge the gap is through H-2A visas and immigration,” he said. “My hunch is that that’s always going to be a necessary part of the equation. But I think that we’re going to see less need for that type of solution. But I don’t think [automation] is going to take a bunch of domestic jobs.”

At Andrew and Williamson Fresh Produce, a large, multi-farm grower-shipper operation that works with, district manager Matt Conroy shares this sentiment. He points to automation as a way to fill what he says is a “10­–20 percent gap in the workforce,” but adds that “our goal is never going to be to get rid of people. At the end of the day, certain jobs may fall to automation. The goal is not to have that happen. But there’s always an uncomfortable reality in there.” A few years back, Andrew and Williamson developed the brand Good Farms in partnership with Costco and the Equitable Food Initiative, a public-private partnership aimed at improving the lives of farmworkers. At its eight Good Farms locations, the company says it includes all its workers in planning meetings, employs them year-round, and provides benefits, among other things. And in an industry known for anonymous disregard at best, and wage theft and sexual harassment at worst, these efforts stand out. And yet Conroy admits that, “at the end of the day, certain jobs may fall to automation. The goal is not to have that happen. But there’s always an uncomfortable reality in there. It’s like the photo booth people—that job went away when everybody went digital.” is the fourth robotics company Conroy has tried working with, and he likes that Cobb and his team are interested in grower feedback, rather than approaching automation purely as a technical problem. And he hopes some of Good Farms’ workers will be able to train to run the automated harvesters, a previously unheard of opportunity in a field that generally offers no opportunity to advance. “It’s about providing more skills to this person now and helping them be marketable in the future, so they can go outside of the scope of just picking from aisle to aisle,” Conroy told me. Of course, while learning to operate the machine on the farm is one thing, really getting trained in the intricacies of the machinery would require that a worker and their family could relocate to Davis for several months—and he said finding that person could be difficult. There has also been an effort to provide a pathway for the children of farmworkers to work in the agtech sector. In 2018, produce giant Taylor Farms invested in two centers, including one in Salinas, where existing workers can learn programming, engineering, and machine operation. And the Western Growers Center for Innovation and Technology—a Salinas agtech incubator—has a partnership with nearby Hartnell community college, where the children of farmworkers have been recruited since 2014 to train for computer science degrees. The idea is to provide a path toward a career in tech without having to leave their families behind.

That promise was one of the things that appealed to Eduardo*, a young man who went through the Hartnell program a few years back. After moving with his four siblings at age 10 from Oaxaca to the U.S. to join his parents, Eduardo (whose name has been changed to protect his identity) spent several years working the fields alongside his family—“cleaning lettuce, cutting onions, stuff like that.” He was good at math and got into the fast-tracked computer science program at Hartnell, but finding a job near his family hasn’t proved possible yet. His first year out of school, Eduardo took an internship for a large ride-share company, learning it was in San Francisco just a few days before it started. Then, the internship turned into a full-time job, and he chose to stay on to learn what he could. He hoped to find work in agtech, but he wasn’t optimistic. “A lot of my friends are jobless,” said Eduardo. “They’re still looking for a job a year or two after graduating. When we spoke last fall, he was eating nearly all his meals in the company cafeteria, living in a surprisingly affordable room with other young tech workers an hour outside of the city, and sending money home to his family—a lifestyle entirely different from most of the other tech workers in San Francisco. “It’s something that I can connect back to my parents,” he added about the prospect of working in agtech. “My dad doesn’t trust getting into a random person’s car. But if I build something for ag, he would trust it.”

Armando Elenes, a farmworker organizer and secretary treasurer of the United Farmworkers (UFW) is skeptical. “They’ve been talking about bringing robots into the field for over a decade,” he said. “I’ll believe it when I see it.” This year, protecting workers from the impacts of the pandemic—and expanding the union’s base, in part so that they have better access to healthcare—are much more pressing issues, said Elenes. But Maria Cardenas disagrees. The executive director of Santa Cruz Community Ventures, and the founder of Undocufund Monterey Bay, was also neck-deep in her work to support undocumented farmworkers impacted by the pandemic when we spoke. She sees the move toward automation as inevitable, and potentially destructive. “Oh, it’s coming,” she told me. “I mean if you look at the millions that are being invested in things like identifying the right strawberry, those millions are not going to go to waste. It’s coming!” In Salinas, as in other agricultural regions of California, Cardenas points to the fact that a whole generation of people have been working seasonal, high-skilled, low-wage jobs in the fields for two to three decades without benefits or any increases in pay. Most domestic farmworkers haven’t had access to much education in the last decade, and in recent years the existing population has been joined by a large number of migrants from Indigenous communities who speak neither Spanish nor English. “Where do they go when these jobs are taken away? There are some young people, but a lot of them are getting close to retirement age, and they have no savings,” said Cardenas. “The employers in ag haven’t really invested in the workforce, and instead treat them as a piece of machinery in the fields.”

In that sense, it’s not surprising that a system that seeks to constantly replace its machinery with a more efficient model would be doing so with human workers as well. But Cardenas adds that most of the farmworkers she and her staff engage with are too mired in the work it takes to survive right now to track the progress of automation—let alone mount a response. “[In Salinas,] they’re in households earning less than $50,000 a year living in a community that takes $94,000 to be okay. So, in many ways, their ability to work is subsidized by community programs and rental assistance,” she said. “And also by informal networks. The family takes care of the kids. You rotate. Or somebody is working in lettuce so they bring lettuce home and somebody is working berries and they bring berries home.” Cardenas sees the value of a response from a union or a community organization, but she hasn’t seen one yet. And while the industry points to the opportunity for better jobs, she’s skeptical about the math, especially because a more efficient harvest won’t likely mean more money for the growers who pay the workers. “All that does is lower your price per pound, with berries in particular. I don’t think a worker who is now running a machine is going to earn so much more . . . to make up for the lost household wages when three people are let go,” she said. And yet, like many in the industry, Cardenas believes it’s likely that the pandemic will speed up the adoption of automation technology in the fields.

COVID has already made life difficult for farmworkers in many ways. “You have a political environment that makes it unsafe for workers to feel like they can get tested or get support. You have overcrowded conditions that makes it hard to isolate,” she said. “And poverty wages, which means that they can’t afford to not work—or access health care. All of that combined is a tsunami, quite frankly. And all of that combined in households that are already living in fear.” Add the demand for produce outside the U.S. and consumer concerns about the stability of the food supply, and the drive to produce will likely take priority over other changes, she says. “Even if the workers are sick, growers will still tend to want to produce. So, I don’t see the time allotted to change the industry,” Cardenas said. “But it will impact the agriculture communities where these workers are living. And the real strain will be felt by cities, which are facing tremendous deficits, and social safety net programs and nonprofits, which are also facing tremendous deficits.”

Samir Doshi, a Race and Technology Fellow at Stanford University, is also concerned about the potential impacts of automation on Latinx immigrant communities in California and he’s engineering a plan to get out ahead of what could be an enormous wave of change. Doshi did his doctoral research on developing regenerative economies for coal mining communities in Appalachia, and sees potential parallels with agriculture. When faced with questions around safety of miners, the companies turned to automation rather than creating safer jobs, says Doshi. When faced with questions around safety of miners, mining companies turned to automation rather than creating safer jobs. “It did make mining much more efficient; it saved a number of costs. And you had mining happening at all times of the day. It basically extracted the value of that industry completely for the owners and operators. You saw a huge drop in employment, and for a lot of mining communities, whether it’s in Appalachia, New Mexico, or other areas, there was no alternative economic pathway for a lot of those communities,” said Doshi. “They didn’t get other jobs within the industry, which is what is being promised in agtech. And they did not move up the career ladder.” Instead, the mining companies, which tended to be based outside the communities where the mining takes place, have all moved their own higher-level employees in to run the machines. Doshi is concerned about this pattern being replicated within agriculture, where immigrant communities play a larger role. “The consequences aren’t just people being put out of jobs. It’s people being pushed out of their homes, their country, their communities,” he said. “It is definitely possible to have dramatically cascading effects on communities and regions for what automation does.”

He has been tracking the rise of agtech outside the U.S., where it’s being funded by many of the same large foundations that have brought genetic engineering to the developed world. Doshi believes that, globally, investment in agtech is “going to be as pervasive as biotechnology,” using a similar narrative of food security and efficiency. Doshi has spent time studying healthcare, education, and other industries that have been radically changed by technology, and hopes to bring a range of stakeholders—from industry representative to academic institutions, foundations, investors, and grassroots community organizers—into a single conversation about how to take workers and communities into account while adopting tech solutions in agriculture. The ultimate goal is a set of principles, or a code of conduct that can help guide the industry. Doshi is looking toward other efforts like UNICEF’s innovation principles and the Digital Impact Alliance, which has created principles for digital development that helped shape investment in the space. “I’m not trying to be predictive. We’re trying to be considerate and strategic about how to take care of our communities, how to take care of our food systems, and how do we look at sustainability and value across both of those domains over time,” said Doshi, who has also worked as a Senior Scientist at USAID, and for the San Mateo Food System Alliance and California Alliance for Family Farms (CAFF) in recent years.

The goal, he says, isn’t to stop technology in its tracks, but to widen the conversation to include workers and smaller farmers who don’t benefit from the same kinds of tech. “There are many technologies that can genuinely benefit small-scale, medium-scale family farms,” he said. “And if we can even the playing field in terms of the utility and efficacy and equity within these technologies and distribute the value that’s provided so that it’s not just large-scale investment into large scale applications that then [only] benefits industrial farming and agriculture.” While the pandemic has slowed down Doshi’s process, he hopes to convene digital conversations about what it would take to develop a code of ethics in agtech—and get buy-in from investors and governments, using mechanisms like the Digital Impact Alliance, the World Economic Forum and other convening agencies and coalitions.

It’s hard to talk about replacing workers with automation without looking squarely at the very real physical cost of farm labor. Flavio Carnejo, a family physician who works with strawberry and raspberry pickers in central California’s Pajaro Valley, described it well in a TEDxFruitvale presentation in 2011, in which he lists the types of pain, swelling, and spasms that occur in the worker’s wrists, shoulders, and backs, as well as longer-term effects, like compression in the sciatic nerve, degenerative joint disease, and arthritis, that farmworkers endure years before most other people do. “Strawberries are picked stooped over, and our bodies are just not designed to do that for so long. You don’t have to be a physical therapist to realize there’s going to be a tremendous amount of damage that’s going to happen to the physical body,” he said. “You touch the back of some farmworkers and they feel like they have rods running up their backs—even years later.”

But it’s not clear that replacing people on the land is the best—or only—way to avoid these problems. And it’s hard not to compare the environmental challenges that come up in agricultural monocrops with the monocrop of movement we see in today’s produce fields. While science points to a diversity of crops as fundamentally better for the environment—it means fewer pests, healthier soil, and cleaner water, for instance—it’s also clear that a diversity of tasks and movement has benefits the human body and brain. “There’s a lack of acknowledgment that the repetitiveness of the motion, which causes physical injuries and then allows for robotic interventions, is really symptomatic of the plantation structure of current agricultural practices,” said Emily Reisman. “Everyone acknowledges that this model is problematic ecologically and socially. And yet somehow we have no choice but to use the plantation to overcome it.”

If the industry weren’t trying to replicate what has been done in factories and other industrial settings on farms, she adds, they may find themselves asking, “what would it take to make economically viable agricultural work that fosters more diversity, or is more intellectually and creatively fulfilling? What would it take to make farm work a pursuit that enriches every life it touches? We know that so many people are desperate for some kind of physical connection to the Earth—not only for their own health, but their psychological well-being, and a sense of place and purpose,” she adds. Meanwhile, Eduardo, the son of farmworkers, is holding out hope that he’ll get to help make Salinas a place where he and his peers in tech can pursue their own sense of purpose. And if automation becomes the norm on Central Valley farms, it won’t be all that different than the changes that drove his parents to the U.S. in the first place. “Immigrants—we don’t do one thing,” he said. “Farm labor is obviously a huge thing we do. But there are a lot of immigrants that work in restaurants and lot of different parts of the world. And we might just have to adapt.”