AI startups want to crack open the recipe book in Big Food's test kitchens

In the realm of large food corporations, the integration of artificial intelligence is becoming increasingly prevalent. McCormick, the company behind popular brands like Frank's RedHot and Old Bay, has harnessed AI in flavor development for nearly a decade. They've reported a significant decrease in development timelines, with projects seeing a reduction of 20% to 25% in time by using AI to identify promising flavor combinations and filter ideas for testing. Similarly, Unilever has embedded AI deeply within its food research and development processes. Their systems can evaluate thousands of recipes in mere seconds, allowing for quicker identification of viable concepts with fewer physical tests. For instance, the Knorr Fast & Flavourful Paste was developed in about half the usual time. On the packaging front, AI has also played a crucial role, as demonstrated by Hellmann's Easy-Out squeeze bottle, where AI helped in modeling formulations, saving significant time in lab work. The journey of AI in food innovation dates back to 2017 when a team from Google Brain collaborated to create a recipe for the 'perfect' chocolate chip cookie. However, even with these advancements, food companies are keen to clarify that AI is not meant to replace human chefs. Annemarie Elberse, who oversees ecosystems, digital, and data for food R&D at Unilever, states, "Human creativity and judgment lead the way, and AI is a tool to help us amplify our impact." McCormick's chief science officer, Anju Rao, echoed this sentiment, emphasizing that AI serves as a co-creation tool rather than a substitute for human skills. "Our greatest asset will always be our people who bring global perspectives, flavor expertise, and human creativity to the table," she explained. As AI startups emerge, aiming to predict consumer responses to new food products using extensive datasets, the effectiveness of their efforts in culinary innovation remains uncertain. Companies like Zucca, Journey Foods, NielsenIQ, and AKA Foods promote their platforms as AI-driven systems that screen recipes, suggest modifications, and forecast consumer preferences before any physical prototypes are crafted. While they promise to streamline the development process, industry analysts point out that many of these startups are still in their infancy, focusing on data collection to build meaningful predictive models. Brian Chau, a food scientist and founder of a consultancy, believes that many AI startups may be overstating their capabilities to attract investors. Current platforms often rely on existing recipes and consumer trends rather than generating innovative products independently. The challenge lies not only in the technology itself but also in the biological complexities of human taste perception. Dr. Julien Delarue, a sensory and consumer science professor at UC Davis, cautions that while AI can enhance efficiency, predicting individual taste preferences is inherently challenging due to the subjective nature of flavor perception. Delarue emphasizes that the variability in how individuals perceive flavors complicates any model's ability to serve as a universal proxy for taste. To truly understand consumer preferences, AI would require access to extensive, individual-level data, which remains a daunting task. Even as AI tools are developed, experts assert that human judgment is irreplaceable. David Sack, founder of AKA Foods, notes that while their platform organizes valuable R&D knowledge, it does not replace the essential roles of food scientists and sensory experts. AI can aid in narrowing down testing options, but human sensory validation remains crucial. The consensus among industry leaders is that large food manufacturers stand to gain the most from AI, as they already possess substantial proprietary data. Dr. Delarue believes that AI's real value lies in enhancing efficiency, enabling faster data analysis, and addressing the complexities of modern food design, which must now consider health, sustainability, and affordability in addition to taste. Ultimately, the human palate remains the definitive judge of flavor. As Jason Cohen, CEO of Simulacra Data, puts it, "Consumers decide with their palate whether they like a product." The journey of AI in food innovation is just beginning, but one thing is clear: human insight will always play a pivotal role in the culinary landscape.