Artificial Intelligence Tools Get Smarter

Artificial intelligence (AI) solutions are reshaping the food and beverage industry—from identifying consumer trends and creating recipes to supply chain optimization and market trend prediction.

The Coca-Cola Company has AI-powered vending machines. Danone tests probiotics with its proprietary “robot stomach.” General Mills has deployed ELF (end-to-end logistics flow) to increase efficiency in its procurement and management operations. And Brightseed’s Forager AI platform helped discover two naturally occurring bioactive compounds that led to Bio Gut Core, a new digestive health product.

AI’s ability to rapidly process information far exceeds human capabilities. The consumer packaged goods industry now stands at a pivotal moment, says Miriam Ueberall, European strategy head for Turing Labs and a member of the Scientific Advisory Board with PeakBridge, a venture capital firm.

“I like it, as long as the human intelligence is not lost along the way,” Ueberall says. She expects AI to enrich rather than replace jobs. She expects those jobs to evolve, however. A future job requirement will be experience in managing AI agents.

“You need to see AI as an extended team member and you need to acknowledge that your skill requirements will change,” Ueberall says. What the industry needs now is a hybrid of a data scientist and a classical food technologist. “They don’t exist yet,” she says.

Ueberall also warns corporate leaders to ensure that they understand how to protect their intellectual property while they are applying AI solutions.

“If you put your request into ChatGPT, the ChatGPT will probably tell you how to create the next-generation chocolate bar, but everybody will get access to that same recipe.” At Turing Labs, she notes, “we never train the general algorithm.” The solution is always customer-specific. “You own the IP (intellectual property). You own the data.”

Ueberall spent 25 years in corporate R&D. Five years ago, in her previous role, she became exposed to Turing Labs after looking to apply AI solutions to improve efficiency, create better outputs, and increase speed to market. Ueberall tested AI against her company’s human product developers. Both her team and AI worked from the same brief and data.

After a few weeks, they compared outcomes. The algorithm reached its result more efficiently and at the same level of accuracy as the R&D team. That showed Ueberall that she could trust the mathematical models.

We are here to advance the profession and be a resource for the profession.

Jay Gilbert

IFT

R&D Gets a Digital Upgrade

“AI can transform the way we do R&D, moving us beyond the slow, manual, people-dependent models that have held the industry back for decades,” she says.

Manmit Shrimali cofounded Turing Labs in 2019, intending to use the technology for ending the trial and error that slows R&D. “But we quickly realized the real challenge wasn’t the models—it was the environment,” he says. “Most R&D teams didn’t have the infrastructure, clean data, or internal management needed to leverage AI. Worse, they had already poured millions into in-house tools and legacy systems that failed to deliver.”

His company’s AI-based digital intelligence platform helps R&D teams develop winning formulations with less guesswork and fewer testing cycles. “It’s not about replacing scientists,” Shrimali says. “It’s about making them more confident and collaborative.”

Traditional work methods such as manual research and focus groups have dominated the food and beverage industry for decades, agrees Alon Chen, cofounder and CEO of Tastewise. Chen notes that traditional product development cycles can take 12 to 18 months but often stall at the insight stage when lacking clear paths to completion.

Tastewise AI simulates sensory profiles by predicting how ingredient combinations affect taste, texture, and consumer appeal. “This allows food and beverage teams to tailor formulations before physical testing, significantly cutting down development costs and time,” Chen says.

In its continuous tracking of consumer sentiment, Tastewise detected the rising appeal of “crunchy” textures in 2025. The data revealed that social conversations about “crunchy” increased by 2.64% over the previous year, and “crunchy” appears on 23.79% of restaurant menus. Tastewise further uncovered “snack” as the fastest-growing consumer crunchy need, and “vegan” ranked as the top crunchy diet.

Amanda Hershon, communications specialist at The Campbell’s Company, says the company uses Tastewise to develop products, content, and more. “I support a lot of our brands across meals and beverages, so it’s really something that I use every single day,” she says in a video posted on Tastewise’s customer stories page. “AI has helped us get closer to our consumers, because we have real-time data that we can access any time.”

Tastewise, based in Tel Aviv, Israel, in 2023 introduced TasteGPT, a sous-chef chatbot to help users understand consumer behavior more rapidly than before. No longer offered as a standalone product, TasteGPT has since evolved into the AI engine behind Tastewise’s generative AI suite for food and beverage brands.

Tastewise AI incorporates consumer insights into product development by analyzing extensive data from social media, recipes, and restaurant menus to spot trending ingredients and flavor pairings. The British supermarket chain Waitrose used real-time Tastewise insights into current and seasonal consumer preferences in developing a Basque cheesecake that became one of its top-selling desserts. In this way, Chen asserts, “brands can lead food trends, not just follow them.”

AI can also help identify alternative sources for ingredients—a welcome capability as companies struggle to cope with rising tariffs. Abhinav Agrawal, partner and managing director at AlixPartners, a global consulting firm, finds that the tariffs have helped spur innovative thinking in AI.

AlixPartners offers a Global Trade Optimizer (GTO) tool that until recently has been used mainly to identify alternative ingredient suppliers. But the tool found new applications for a leading beverage company after the recent tariffs took effect.

The company had been using monk fruit, which comes from China, as a zero-calorie sweetener. When the tariffs suddenly raised the cost of monk fruit by 25%, the AI-based GTO recommended stevia as a substitute. That option was cost-effective, Agrawal says, because “you can import stevia from India, where the tariff was 10%.”

The tool also suggested some formulation options to help mitigate stevia’s aftertaste. Humans still need to make the decisions and carry out testing, Agrawal observes, “but very quickly, the sourcing could be adjusted.” GTO also suggested using sucralose, not subject to tariffs, in combination with other ingredients as another zero-calorie sweetener option.

Increased computer capacity, the emergence of ChatGPT, and lower AI platform costs have all contributed to the AI trend in product development. No longer limited to large companies, even startups are adopting AI now, Agrawal observes.

The process still requires human participation, but fewer people are needed and the results come faster. “You can use the speed to drive more products or to drive organizational efficiency. It’s the company’s choice,” Agrawal says.ft

New IFT AI Platform Powers Innovation 

IFT unveiled its new artificial intelligence platform, called CoDeveloper, at IFT FIRST Annual Event and Expo last month in Chicago. A system built by food scientists, for food scientists, the platform includes a generative AI system called “Sous,” which leverages the same technology that powers ChatGPT to help accelerate R&D through science.

“You can see from the name CoDeveloper that this is a companion meant to support you in the development process and not to replace the work you’re doing,” says Jay Gilbert, IFT’s director of scientific programs and product development. CoDeveloper’s chat feature allows users to ask Sous technical questions and learn how the system works.

“CoDeveloper allows us to deliver scientific knowledge and information into the product development and R&D process, and to be more accessible to developers when they need it,” Gilbert says.

The platform supports the early-stage development process from ideation to benchtop formula management, streamlining R&D and rapidly overcoming technical problems. With CoDeveloper, when food developers receive a project brief from their company’s marketing team, they can start the formulation process with a focus on their unique requirements. Additional features include computational formula development, reverse engineering, and adjustment and assembling ingredients from a proprietary catalog.

Security is a key feature of the system. “The questions you ask Sous, the formulas you are iterating, the ingredients you’re using within our system are all yours. We are not using that to train our data or our algorithm,” Gilbert says.

As a retrieval-augmented generation system (RAG), CoDeveloper combines information retrieval with generative AI to produce its responses. Gilbert’s team took all of IFT’s content and made it accessible to the RAG system’s large language models. These models then generate a response and connect users to the original source for expanded learning.

“When a question comes in from a user, Sous finds the right piece of information within IFT’s content and brings it back to the user,” Gilbert says. He likens the system to an Amazon warehouse that contains all of IFT’s content. ChatGPT already exists, Gilbert notes, but how many questions do users have to ask to get the information they seek? “Your prompt has to be perfect for you to get the right information. Our goal is to do that for you.”

Early versions of large language models often suffered from AI hallucinations that resulted in confidently expressed false or inaccurate information. Gilbert’s team, however, built the IFT system to leverage the most advanced language model from OpenAI. This enables the system to understand the questions asked. CoDeveloper’s responses include links to its source references, should users find the information relevant.

“We are here to advance the profession and be a resource for the profession,” Gilbert says. “This is going to be such a transformative technology that we have a unique obligation to make this technology accessible.”

AI is nothing to be afraid of, he asserts. “My mantra is, AI is not going to take your job, but somebody with AI will.”

Learn more at codeveloper.ift.org.