For years, packaged food companies and the agricultural manufacturers they work with have used AI to help increase crop yields and create more desirable food formulations for consumers. Now, the food-manufacturing industry has a new AI tool for boosting productivity on farms and in factories: advanced large language models.
With this type of generative AI, food companies can pull together disparate pieces of information — like alerts for global tariffs, expected fungal outbreaks that may require pesticides, or strong winds that could affect moisture levels — and use that data to make more informed decisions about growing and buying key ingredients.
The global food production industry, which is worth an estimated $4 trillion, has the potential to generate $250 billion in annual profits from AI’s productivity potential, as per a 2024 McKinsey & Company report. These potential savings from more targeted labor and greater operational efficiencies in manufacturing come at a critical time, when global food commodity prices have increased to their highest level in two years in July, as per the United Nations’ Food and Agriculture Organization.
For effective implementation, leaders must contend with how to apply AI to complex agricultural systems. Challenges include recruiting engineers, software developers, and data experts and organizing data uniformly across the supply chain — from small family-owned farms, which may have limited resources, to agricultural producers and large national retail chains, as per the McKinsey & Company report.
Despite these challenges, big food manufacturers — like Land’O Lakes, PepsiCo, and the global agricultural service provider Cargill — are implementing AI to improve farm yields, boost the productivity of food manufacturing plant workers, and ensure deliveries of products like butter, milk, and protein cereal meet anticipated retailer demand.
Cargill’s AI tools make deliveries to Walmart more efficient
Cargill uses an AI computer vision tool called CarVe, which detects how much beef the company’s workers remove from any given animal carcass, said Jennifer Hartsock, the company’s chief information and digital officer.
If too much meat is left behind, CarVe flags it and shares those insights with Cargill’s shift managers, who can retrain workers to get more precise with their knife skills.
“It’s a very expensive commodity in the market, and we won’t want waste to get sent down the stream and out the back of the plant,” Hartsock told Business Insider. Any loss of meat would result in a more wasteful supply chain and raise costs for Cargill and the consumers buying its products, especially when ground beef prices have recently hit record highs.
Further down the supply chain, Walmart shares its sales data with Cargill, said Hartstock. Cargill then uses AI to analyze the data and generate production recommendations. If there are swings in demand for foods like ground beef or London broil, for example, Cargill factories can adjust their plans quickly to ensure shelves stay full while avoiding costly surpluses.
AI helps Land O’Lakes plan for dairy-demand spikes
This year, Land O’Lakes — which operates both a farm business and a dairy business — debuted a generative AI tool in partnership with Microsoft. Agronomists can use the tool to help farmers make more data-informed decisions about crop production and soil management.
According to the company, agricultural experts can input details about the farms they are visiting — including the time of year, weather, the type and amount of soil used, and the maturity of the crop — and get AI-generated suggestions for making a particular farm more productive without increasing costs.
Another application of AI involves demand prediction for the company’s dairy business. Land O’Lakes works with nearly 1,300 dairy producers across the US. The cows on those farms produce milk at consistent levels throughout the year, but demand for Land O’Lakes’ dairy products like butter peaks during holidays like Christmas. This creates an imbalance in food production and sales, said Teddy Bekele, the chief technology officer at Land O’Lakes.
“You can’t go to the cows and say, ‘It’s game time, let’s produce as much as we can,'” said Bekele. “They are going to do the same thing they do every day.”
To help predict these kinds of demand fluctuations, Land O’Lakes uses AI to flag when larger amounts of Land O’Lakes-branded butter will likely be in demand, or when the company should instead focus on selling milk in retail stores.
Predictive AI helped PepsiCo create new high-protein oats
Over the past two years, PepsiCo used AI’s predictive capabilities to aid in the creation of new oat varieties that contain more protein, according to Ian Puddephat, PepsiCo’s vice president for ingredients. This allows the company to develop and sell Quaker Oats products that protein-hungry consumers are increasingly seeking.
Using AI in this manner also has a positive environmental impact: Before growing oats that are naturally high in protein, PepsiCo would boost oat crops’ protein levels with whey, a milk byproduct that typically produces a higher environmental footprint than standard oats, Puddephat said.
He added that an AI algorithm helps PepsiCo predict which two parent lines of a plant would be best to cross-breed to create varieties that use less water and land and require less fertilizer or agricultural chemicals compared to prior generations of those plants.
