By Isabelle Bousquette
Kevin Wallenstein is a health nut, according to his colleagues at snack behemoth Mondelez International. He runs and paddleboards, and when he does snack, it's on fruit, not the Chips Ahoy and Oreos that Mondelez makes.
So working in biscuit digital research & development -- where cookie and cake tastings may be necessary four to five times a week -- can be trying, said Wallenstein, who is section manager.
"The number of tastings we have is not fun," he said. "I used to work in Sour Patch Kids, and if you did a tasting every day for a week, it was a nightmare."
But now artificial intelligence could transform Wallenstein's life. A new tool developed by Mondelez is speeding up the creation of snack recipes and optimizing them to fit certain taste profiles. Which means less lab work, faster time to production and -- for better or worse -- fewer in-house tastings.
"The point is we get there faster," Wallenstein said. "The consumer wants the product to taste like X. We're not stopping iterating until it tastes like X ... we're doing things more efficiently."
Food companies like Mondelez are racing to try out AI in every area of their business, from supply chains to marketing to recipe development, said Marion Nestle, professor emerita of nutrition, food studies and public health at New York University.
The companies are under pressure: Capturing the American public's attention is no small feat in a jam-packed foodscape that churns out more than double the amount of recommended calories for each person in the country. New types of products and brand offshoots are the way to do that, Nestle said, and if AI can get there faster and better, expect to see its role grow.
Not everything will change though. "Seems to me you still need human tasters," Nestle noted. (Sorry, Kevin).
Besides Oreos and Chips Ahoy, Chicago-based Mondelez makes Ritz crackers, Cadbury chocolate, Clif bars, Triscuits, Tate's cookies and Toblerone, among its many well-known nibbles. It reportedly approached Hershey's about a merger recently, but the chocolate maker presents a challenging takeover target and has rebuffed Mondelez in the past. While chocolate Kisses may not be coming to Mondelez's portfolio anytime soon, the company will be updating its classics with an assist from AI.
Mondelez's North America R&D center is based in a pinwheel-shaped building in rural New Jersey, where snack mascots abound and almost everything is colored signature Mondelez purple (including the receptionist's hair). Food scientists there use the AI tool to create optimal recipes by specifying desired characteristics, including flavor ("buttery," "in-mouth saltiness," or "vanilla intensity," for instance), aroma ("oily," "egg flavor," "burnt," among others) and appearance ("amount of chips," "roundness," "chip edges" are considerations). The tool also considers parameters like the cost of ingredients, their environmental impact and their nutritional profile.
Employees are extremely tight-lipped on Mondelez's proprietary recipes, but the AI tool does work with that information. It then suggests a handful of new recipes that meet the requirements, a process that previously was more or less trial and error. To be sure, the AI suggestions are taste-tested, but new products are getting to pilot or production trials four to fives times faster, the company said.
The tool began development in 2019 with software consulting firm Fourkind, later acquired by Thoughtworks. It uses machine-learning, a more traditional form of AI, rather than the generative AI popularized by ChatGPT. Thoughtworks engineers do most of the coding, but they work closely with Mondelez data scientists to ensure the tool is trained on plenty of snack-specific information.
Now Mondelez says use of the tool has reached critical mass for biscuits and baked snacks in North America and for powdered beverages globally. It has been used in over 70 different projects.
The results range from brand offshoots like the Gluten Free Golden Oreo to refreshed recipes on classics. No, Chips Ahoy haven't always had exactly the same recipe: Existing products are assessed every few years for potential tweaks, like whether to move over to a new chocolate-chip or vanilla-extract supplier.
Mondelez has weathered tough economic conditions over the last couple of years slightly better than its competitors, in part because of its ability to iterate quickly and adapt to consumer behavior, said Jefferies analyst Rob Dickerson. The company's organic sales in the latest quarter grew 5.4% from a year earlier.
Mondelez Chief Executive Officer Dirk Van de Put is known for pushing constant investment in the company's brands and products, even when the market is ailing, Dickerson added.
But there are limits to how far Mondelez will allow its AI to go. Consumers tend to respond to new products that aren't overly unfamiliar, which is why brand extensions are so popular, said Nestle, the nutrition expert.
Food scientists using the tool do select different characteristics, but work closely with "brand stewards" of each snack to make sure that basic requirements are met.
Earlier iterations of the tool that weren't given as much data made some unhinged suggestions.
"Because [baking soda is] a very low-cost ingredient, it would try to just make cookies that were very high in baking soda, which doesn't taste good at all," said Wallenstein, the biscuit digital R&D section manager.
The brand steward is key, Wallenstein said. "They'd be telling the tool: This is what the essence of an Oreo is."
So how might the essence of an Oreo gently evolve? Well, that's between Mondelez and its AI.
Write to Isabelle Bousquette at isabelle.bousquette@wsj.com
(END) Dow Jones Newswires
December 17, 2024 07:00 ET (12:00 GMT)
Copyright (c) 2024 Dow Jones & Company, Inc.
Comments