Data's role in AI-fueled product development

BACK TO ALL NEWS

Data’s role in AI-fueled product development

3 Nov 2025

Mari Rydings

Mari Rydings is a versatile content creator, editor and marketer with experience developing dynamic, results-driven materials for B2B and B2C audiences. As the editorial director for Avant Food Media, she leads the day-to-day editorial operations for the company’s print and digital magazines, newsletters and social media platforms. Outside of work, she loves reading, hiking, putting together jigsaw puzzles and spending time with her husband and their twin daughters.

KANSAS CITY, MO — As a testament to AI’s increasing use in recipe formulation, the Institute of Food Technologists (IFT) launched its AI-powered R&D platform, CoDeveloper, at this summer’s IFT FIRST event. Designed to speed up formula development and help food scientists with ingredient substitutes, texture improvement, clean-label reformulation and flavor development, the platform features a generative AI co-scientist, advanced formulation tools for iterative development and science-optimized reverse engineering.

While AI tools have value in individual areas of new product development, their strength lies in cross-functional application. When Chicago-based Mondelez International decided to innovate on its Chips Ahoy! cookie, the company turned to AI for help. Project leaders took steps to ensure team-wide adoption of the new approach to product development.

“We encouraged people to think of it as an insights source versus a data source,” said Maria Rondon, director of brand insights, confections for Mondelez International, during a recent Industry Dive webinar. “At Mondelez, AI is truly cross-functional, especially as we think about new innovation pipeline development. With Chips Ahoy!, we brought the entire team along on the journey: consumer insights, marketing, the innovation team, consumer science, R&D. These partners saw the value that this new process was bringing to their specific functions.”

The power of any AI tool lies in the data it receives. The more accurate information it gains, the smarter the AI becomes.

“If you want to be the best partner for your customer, you need the best tools and the most insightful data,” said Benjamin Bado, data and digital innovation manager at Puratos. “A company’s true intellectual property isn’t its machines. It’s the data and knowledge the company has accumulated over time. There must be something unique about the data you possess, because if you build the same tool with the same data source, anyone can replicate what you do. You want your own recipe, your own knowledge and your own method for creating it.”

“There must be something unique about the data you possess … You want your own recipe, your own knowledge and your own method for creating it.” — Benjamin Bado | data and digital innovation manager | Puratos

To maximize the benefits of an AI model, companies must make sure their data is clear, concise and organized into the right model for the right position. Dissecting existing documentation into specific knowledge sections and uploading each section into the appropriate AI model will be more effective, but it can also be a very time consuming and arduous process.

“One thing I see in the future for ingredient suppliers and equipment manufacturers is customers asking for all the information inside a typical manual — operator settings, changeover procedures, maintenance tasks, sanitation tasks, troubleshooting, diagrams — to be organized into individual documents for specific audiences such as production, R&D, maintenance and food safety,” said Justus Larson, VP of operations for Portland, OR-based United States Bakery during the BEMA BIF panel.

That’s where things can get tricky. According to the American Bakers Association’s survey, 38% of companies reported that the availability of internal data for use with AI was one of the top five challenges to implementing the technology. It’s not just a lack of data that impedes AI adoption. It’s also the increased risk exposing proprietary data, including formulas, to competitors, with 65% of survey respondents indicating cybersecurity was the primary AI-related risk most relevant to their businesses.

The growing concern of a data breach of any type has prompted many companies to construct sophisticated network firewalls that restrict communication between secure internal systems and less secure external networks, and that makes AI implementation tough, but not impossible.

“We are locked down like Fort Knox,” said Richard Ybarra, senior manager of manufacturing engineering for Lakeland, FL-based Publix, during the BEMA panel. “When I started talking to my leadership about implementing AI in our facilities, the first thing they said was that we couldn’t go outside of our firewall to pull data. That was a challenge. But we went back and looked at how we could incorporate AI in various areas within this, and other, limitations.”

The use of AI in the food and beverage industry will only accelerate the pace of new product launches. Companies that recognize its value and embrace its capabilities will find new opportunities to excite consumers, a unique competitive edge and the freedom to ask, “What if?” … pushing the boundaries of taste, texture, nutrition and flavor.

This has been adapted from the August | Q3 2025 issue of Commercial Baking. Read the full digital edition here.

Advertisement

BACK TO ALL NEWS