Sniffing Out Particles, Singling Out Pigs: Food Industry Urged to Embrace Robots, AI

E.coli in beef. Metal shavings in bread crumbs. Contaminated water on vegetables. The food industry has its share of quality control scandals and their insurers have had their share of claims.

Some believe food would be a lot safer to both eat and insure if humans were less involved.

“Every year there are deaths and there’s significant financial damage to the overall industry as a result of these situations. One of the biggest challenges we face in this whole space is that the entire market is driven by human error. Accidents happen, food processing facilities are messy places,” according to Mark Leblanc, head of Crisis Management, North America, for Swiss Re Corporate Solutions.

At this year’s Risk and Insurance Management Society (RIMS) meeting in Boston, Leblanc demonstrated how robots and artificial intelligence can do a better job than mere humans of keeping the food supply safe.

Leblanc, whose unit at Swiss Re handles product contamination and product recall insurance, said one reason quality control is perhaps not what it should be is because the food industry tends to be occupied with other business menu items. His unit also deals with auto manufacturing risks and Leblanc said the investment in technology by the food industry doesn’t come close to that in the auto industry.

“A lot of times the companies that buy this [food contamination and recall] insurance are in the position where they’ve got tight margins, they’ve got tight timeframes, and they’re in a situation where they’ve got to wear multiple hats,” he said. “They’re dealing with quality control issues, they’re dealing with food safety issues, they’re dealing with finance issues, and as a result, resources are spread thin.”

He said companies can be distracted and, as a result, they have limited financial resources available to invest in quality control and food safety.”

What the technology can do varies by industry.

“The robots on the assembly line for an automotive plant, they aren’t dealing with the same issues as the food sector. The food sector has bacteria, it has allergens, it has whey, it has incoming materials. It has a lot of other issues, where just an automated processing or assembly line isn’t enough to actually reduce the risks,” Leblanc said.

While automation is not currently prevalent within the food industry, some segments are now investing in robots and artificial intelligence, according to Leblanc.

For example, some food manufacturers are now using robots in kitchens to make prepared meals. A robot can complete 100 to 200 meals on its own before needing to be refilled.

Robots are also being used for cutting and sorting chicken and beef. “They use a robot to separate parts of the meat. You can see this stuff utilized in rendering facilities where you’ve using different parts of the chicken,” Leblanc said

AI and robots are also being utilized in detecting microscopic bacteria as well as foreign metal and non-metallic materials such as vinyl, plastic and cardboard. “It doesn’t come up on x-rays, or on metal detectors,” he noted.

The new technologies are also used by processing plants on assembly lines to pick and properly label containers, while also verifying that the correct label is being used so that people with allergies are properly alerted. AI and robots can check 200 product labels a minute, which adds up to a lot of time saved in a massive food plant over error-prone humans doing the checking.

“What you have a lot of times in this industry is with all the allergens that are out there in basically the community and society and the environment, you have a real exposure for companies that may mislabel and forget to point out that an allergen may be in the product,” he said.

Water quality is a growing area of concern where there is potential for the application of the new technologies, said Leblanc, citing past scares around romaine lettuce and spinach. “They were never able to determine the source of the contamination until they checked water supplies,” he said.

“If you think about leafy greens, that comes out of the ground, it gets washed, it goes straight into the packaging, straight on your shelves at the grocery store,” he told the RIMS audience. “The only real interaction with that product is the rinse and the packaging, and in some cases, there is no packaging. The ability to quickly identify if you have issues in your water supply is a very important tool that, frankly speaking, needs to be better utilized and applied more significantly in the food processing industry.”

AI is being used as a way to identify contaminants by both sight and smell.

Large pork processing plants in China have begun using facial recognition for their pigs to be able to trace every pig through the facility. This can help identify the one animal in a massive population that has avian flu or some problem. “If you have a chicken farm and one of the chickens has avian flu, you need to be able to identify the chickens and trace the chickens,” he said, noting that humans can’t tell them apart but AI can do so quickly with its heightened vision capabilities.

Farmers in the Midwest are “terrified” of the possibility of African swine flu or avian flu, Leblanc said. “It’s the ability to trace, identify and, if need be, separate some of these animals that may have these contaminants that’s critical. Ultimately, these illnesses can make their way into the food chain,” he said.

AI that can smell, sometimes called an electronic nose, can sniff out “incredibly tiny and minute particles, parts per million, parts per billion type of particles” that could make their way into water and products.

According to the Swiss Re food industry expert, AI is better than the sample testing now being used to determine if large batches of ingredients or products are safe. He used the example of a silo full of flour. A farmer or processor is not going to sample every inch of that silo. Certain contaminants may not spread throughout the entire sample size. “It might be literally a needle in a haystack. The ability for AI to sift through this stuff and determine where that actual pathogen or bacteria may be is incredibly impressive and powerful, and frankly, needed in the industry,” he said.

AI is also being used to track and analyze social media, where the concern is adverse publicity over a food product said to have gone bad. Analyzing social media helps trace where a product has been and where it is going.

“What they’re beginning to use if for is a way to trace where the product’s going, people’s reaction to that product, so that they can nip the issue in the bud, and if need be, the company can proactively intervene and either stop the product from further distribution, contact customers, do a press release, and essentially cut it off at its knees before it gets out of control,” he said.

Leblanc cited a case of Amazon pulling an energy drink off the market after someone wrote on social media, “Gave me energy for my workouts but also wreaked havoc on my digestive tract.” He said Amazon was tracking social media and was thus able to intervene quickly.

Going a step further, the AI tools are also being used to track purchase patterns and who has bought what, when and where. They can also be used to advise purchasers if a problem with a product is discovered. “You’re seeing it used not only in identifying contaminants, but you’re also seeing it utilized in traceability of where the product’s going, what the trends are with the product and who’s using the product,” he said.

There are contamination scenarios that transpire over long periods of time such as when people in Chipotle restaurants in an area were getting sick. Chipotle was having difficulty understanding why and where the contamination was occurring and how to trace back. In this situation they actually used whole gene sequencing to go back and find the source of the issue.

A major effect of using automation in the food industry is it replaces jobs. “Yes, these are taking people’s jobs and we understand that, but at the same time what it’s also doing is significantly minimizing the chance for human error,” Leblanc said.

Among the workers at risk are those who clean the food processing plants— more plants are utilizing automation to sterilize their equipment and work areas. “These third parties are humans, and they make mistakes. …,” Leblanc continued. “We’ve seen losses in the industry where the sanitization of the plant isn’t carried out correctly. There are contaminants left in the plant. The next day, they assume it’s sanitized, they go into production, it turns out it’s not.”

He said his own firm currently has two losses with this type of sanitation situation.