Nearly three years ago, when Unilever developed a zero-salt bouillon cube for its Knorr brand, it needed to design a new formulation that didn't contain key ingredients that contribute significantly to taste and give important structure to dried soups. there were.
To predict the optimal combination and concentration of ingredients for a new salt-free stock cube, Unilever turned to artificial intelligence.
“Predictive modeling can be very helpful when creating bouillon cubes that are salt-free while still maintaining a taste that consumers will accept,” says Carla Hilhorst, chief research and development officer for Unilever's nutrition business. states. AI helped design the formulation of the salt-free bouillon Her Cube and how it is made on the company's production line.
The world's largest food companies have relied on AI for decades and have accelerated technology adoption in recent years, including new use cases for generative AI. “CPG [consumer packaged goods] Michelle McGuire Christian, Chief Commercial Officer of ConvergeConsumer, a joint initiative between Deloitte and Google Cloud. “They have been using this to optimize their supply chain for a long time,” she added. This also includes merchandising and product placement within the store.
Unilever's AI models are used to predict taste, consumer preference, microbiological stability, and determine whether a product can run on the factory line. Using AI, the time required for product development is reduced from months to just days. “Data-driven decision-making allows us to find more of the sweet spot,” says Hilhorst. “I can be more precise.”
Other AI innovations include Hellmann's vegan mayonnaise. Unilever used advanced modeling to predict the taste, texture and stability of its egg-free products. AI helped Unilever reduce food waste. Hellmann's squeeze bottles now have a thin coating of oil on the inside of the package so that only one to two servings of mayonnaise remain in the container, compared to an average of about five servings before the AI formulation prediction model was utilized. has been added.
AI also helped Unilever pivot after Russia invaded Ukraine and the company needed to quickly formulate alternative formulations for hard-to-source ingredients, such as sunflower oil.
Athina Kaniura, PepsiCo's chief strategy and transformation officer, said the snacking giant is moving away from more “traditional” formats, from innovation to planning with suppliers to the best routes drivers should take to deliver products. He said he has been using AI for many years. The company has also begun testing generated AI in several parts of PepsiCo's business, including software development, call center inquiry handling, and food formulation.
PepsiCo has developed internal standards to ensure that everyone using AI can use it safely. “We have an infrastructure in our environment that allows us to access an AI sandbox and test different features,” he says.
And as the world of snacks constantly evolves, AI can help synthesize what consumers want at a faster pace, while also resolving the best path to developing those foods and beverages across the supply chain. PepsiCo says it can. PepsiCo emphasizes that human insight is at the heart of these flavor development decisions. “The next big thing is not going to come from technology, it’s going to come from the consumer,” Kanioura says.
PepsiCo says its AI efforts are focused on efficiency and growth, not expenses. “We are in no rush to optimize costs with genetic AI,” he says.
“AI helps us achieve unprecedented levels of accuracy and speed,” said Anton Vincent, president of Mars Wrigley North America. “As we start thinking about removing complexity, we expect AI to be a big driver for us.”
The candy maker says it has been using AI for several years to help source, manufacture and digest industry data to determine trends in the food industry. Regarding large-scale language models, Vincent says: We're still in the early stages of that. ”
To advance that journey, Mars will open an AI lab at its Newark headquarters in the first quarter of 2024. “The AI Lab provides an opportunity for every employee to come in with an idea and hopefully leave with a plan of action,” said Gabriel Wesley, chief marketing officer at Mars Wrigley North America. Masu.
Marie Wright, chief global flavorist at food processing company ADM, recalls the days when flavorists would write down in “beautiful” books the formulas that would form the basis of flavors produced in factories. However, the move to computers, and more recently her move to AI, has changed the way the industry processes formulations and data.
“AI platforms have come a long way in the last few years,” Wright says. “And how can you apply that to creative skills like flavor creation?” She says, “Many flavorists are afraid of AI. And most creative people are afraid of AI. I think you're afraid of that.”
ADM says it's not using AI to take away flavor development work, but to improve the process. Wright says it's important that technology groups don't just mandate the use of AI tools. Instead, the flavorist should be encouraged to join his AI journey.
“We humans can't process that data, so AI has to come into play,” Wright says. “AI has the huge advantage of being able to quickly process and learn from that data, and eventually start doing machine learning to create new formulas, new ways of working, and perhaps even more predictive things in my field. The hope is that we will be able to do it, and the extrapolation is that we should be better at creativity.”
Bayer says it was a fairly early adopter of AI tools such as machine learning. Let's take the example of corn. Before plant breeding, approximately 20 to 30 bushels were grown on one acre of land. Currently, the average corn yield is 175 bushels. Some of these gains were due to farms evolving fertilizers and farming methods, but most were due to genetic improvements made by AI.
AI will help Bayer process data to determine the best parental candidates that breeders can cross to produce the most successful corn offspring in the field. There are billions of possible combinations of corn genes that contribute to yield.
“It is clear that the human mind is not capable of absorbing and fully understanding all the vast amounts of data we are currently collecting,” said Bob Reiter, global head of research and development for crop science at Bayer. “I did,” he says. “Being able to break through this and unlock even more genetic potential that was previously not possible without AI tools is huge.”
Predictive algorithms about which parents to mate and which genetic combinations to create are so sophisticated that Bayer now only has to do it three to four times a year, a process that used to happen every few years before the advent of AI. Now you can do it twice. “This will be a huge driver in the future to continue increasing farmers' productivity of their planted acres,” Reiter said.
Additionally, with the world's population expected to reach 8.5 billion in 2030 and rise to 9.7 billion in 2050, farmers will need to become increasingly efficient.
“This is the basis for feeding a growing world population,” says Reiter.