AI Is Helping Automate One Of The World’s Most Gruesome Jobs

But for all the strides processors of beef, pork and chicken have made bringing automation into their factories, mechanizing the tasks with the most direct impact on profits—slicing carcasses—has been much slower.

That’s because in the messy business of disassembling livestock, leaving even a fraction of an ounce of meat on the bone can erode margins over time. Although robots don’t fall victim to injuries or illness, as people can, humans possess an ability to make on-the-fly decisions for each uniquely sized cow, chicken or hog coming down the line. That’s traditionally given them the upper hand in precision butchery.

“Our challenge is we can’t deploy automation in jobs that would then compromise yield because, given the volume we produce, our operational efficiency really drops if we do that,” says Chetan Kapoor, who oversees automation at Tyson Foods Inc., which processes almost 40 million chickens a week.

That may be about to change. Thanks to recent advancements in machine learning, computer vision and artificial intelligence, the hardest-to-automate tasks, such as carving the perfect chicken breast or meticulously breaking down massive hogs or cows, are finally within reach. If applied on a commercial scale, the technology has the potential to ease the industry’s long-standing labor vulnerability—fully on display during the Covid-19 pandemic, when employee absenteeism contributed to bare supermarket shelves.

Tyson says it has developed a sensor- and camera-equipped robotic chicken breast deboner that’s more efficient than the average worker using their hands; it’s a high-margin task the industry had previously struggled to entrust fully to robots. The machine, which has been in development since 2015, is “way beyond the testing phase,” Kapoor says, showing a video of it in action during a tour of Tyson’s automation center in Springdale, Arkansas. “We took out the variability caused by a very-difficult-to-staff job in a very critical part of the business.” Plus it leaves less meat on the bone, he says.

Automation has long been on processors’ wish lists. Their plants are cold and humid with a high turnover rate, making staffing a persistent challenge. But raising productivity became especially important during the pandemic, when the industry was slammed with a severe shortage of workers, many of whom fell sick on the job while laboring in close quarters. More recently, the US sector has been hit by a supply glut and high animal feed costs, which have highlighted the need to reduce expenses and boost margins.

At Wayne-Sanderson Farms, a poultry joint venture between Cargill and Continental Grain Co., robots now perform almost two-thirds of all chicken deboning, more than twice the rate before the pandemic, says Chief Operating Officer Kevin McDaniel. “When Covid hit and we couldn’t get the labor, we were forced into it just to meet customer demand,” he says. Machines are now “equal to, if not better” than handwork at deboning dark chicken meat such as thighs, as “technology has gotten better and better.” Still, when it comes to white-meat deboning, McDaniel says, “you can’t beat people doing it by hand, still.”

Companies say automation doesn’t necessarily mean a big reduction in head count; instead, they can reassign workers to other roles such as inspection or, frequently, just plug holes where workers have quit. Smithfield Foods, the largest US pork producer, this year expects to spend more than double what it invested in automation in 2021. Smithfield, owned by Hong Kong-based WH Group Ltd., has reduced its factory floor operational workforce 6% over the past three years, helping the company weather its high turnover rates, says Keller Watts, its chief business officer. One of the new tools Smithfield rolled out was a robot capable of scanning carcasses and deftly pulling the rib from the loin, one of the most strenuous jobs for human workers.

“We need to go ahead and get on board because we know the labor force that existed 20 years ago is not the labor force that exists today,” says Brent Glasgow, who manages a Wayne-Sanderson chicken facility in Tyler, Texas, where robots help process 1.3 million birds a week. “And the labor force 20 years from now is going to be even different than it is today.” Glasgow says some chicken processors installed deboners 15 or 20 years ago but ultimately reverted to human labor for some of the harder tasks, because “the technology was not there.”

Automation has been much slower in pork and beef butchery, where the size of animals can vary by tens or hundreds of pounds, making the breakdown almost surgical in nature. Unlike car manufacturing, where almost every product to roll off the line is identical, “every animal is a little bit different, so you cannot standardize the disassembly process,” says Hans Kabat, president of Cargill’s North American protein business. “You have to have sensors—including vision, feel and flexible fingers—on robots to be able to sense where things need to get done and then to actually move the product.”

Only a very small percentage of beef processing can be automated because the technology is either not ready or not scalable, says Enrique Villars, manufacturing excellence leader for Cargill’s North American protein business. But that could change fast as the tech improves. At its beef plant in Dodge City, Kansas, for instance, Cargill has been able to install saw lines that separate meat from bone using a 3D vision system, eliminating the need for workers to make as many as 3,000 cuts a day with a band saw. AI also detects any foreign objects on the production line.

“Computer vision is the next big thing, because it is the ability of the computer to interpret the images and pixels and translate that into real actions,” Villars says. “The speed in which things are progressing is exponential.”

Meat processors are putting more capital into technology projects. In 2021, Tyson said it would invest $1.3 billion in automation over the next three years, including building a highly automated chicken nugget factory in Danville, Virginia. Chief Executive Officer Donnie King says the plant can produce as much as 30% more product with a third less staffing than a sister plant without the same level of automation. Cargill has completed 114 automation projects and has more than 120 others still underway across its meat operations as part of a $700 million plan dubbed “Factory of the Future.”

“You may go to a poultry operation that seems highly automated, and then you walk into a pork or beef operation, and that is maybe underwhelming,” says Smithfield Foods’ Watts. “I think it will look vastly different in five years.”

More stories like this are available on bloomberg.com

©2024 Bloomberg L.P.