How AI Is Rewriting the Food Innovation Timeline

At the 2025 IFT FIRST Annual Event and Expo’s Hot Topic Studio session Anticipating Appetite, Zak Stambor, senior analyst for retail and e-commerce at eMarketer, spoke with Oisin Hanrahan, CEO and co-founder of Keychain, about how AI is changing product innovation cycles across the food and beverage sector. Hanrahan, who previously spent two decades building technology companies, described how AI-driven speed is reshaping how brands identify opportunities, iterate on ideas, and compete in an increasingly crowded market.

Stambor: You have a long technology background. What brought you to the food and beverage space?
Hanrahan: I’ve spent many years building technology companies. What drew me into this space was seeing a consistent challenge: companies of all sizes struggle to identify who can make a product, what processes are needed, and how to move from idea to production efficiently. There isn’t much information flow across the industry, and that slows timelines and creates unnecessary barriers.

Stambor: You’ve said that companies winning today aren’t reacting—they’re anticipating. What makes anticipation so important now?
Hanrahan: The landscape has shifted. Several years ago, much of private-label activity involved replicating successful products at a lower price. Today, retailers are using their consumer and trend data to develop unique products designed to build loyalty. That means the competitor who once followed your lead may now move ahead of you.

At the same time, influencer-driven brands can scale quickly because they start with an audience already in place. Traditional brands face pressure from both directions, so having a unique product or packaging approach—and being able to move quickly—matters more than ever.

Stambor: How is AI changing innovation cycles?
Hanrahan: AI has made early development steps significantly faster. One way to use it is simply to take advantage of existing tools—drafting concepts, structuring documents, or generating starting-point materials. Another approach is integrating an organization’s proprietary data with these tools. Large companies with customer insights, sensory data, or product-performance information can combine those assets with enterprise large language model environments to generate insights that others can’t.

Smaller brands don’t have the same data, but they can move quickly. Large organizations may need months to align on data governance, approvals, and compliance. Smaller teams can take multiple “shots on goal” during that time. AI levels the field by removing many of the delays that used to slow early-stage development.

Stambor: Can you give an example of how speed has changed?
Hanrahan: Steps that previously took weeks—developing a spec sheet, outlining ingredients, drafting packaging directions, or identifying potential manufacturing partners—can now be completed in minutes. We’ve seen teams go from an initial idea to concepts they can share internally or externally almost immediately. In several cases, that acceleration has allowed teams to move from concept to physical samples in roughly 45 days. That same process used to take close to six months.

Stambor: You’ve emphasized proprietary data as a long-term differentiator. Why?
Hanrahan: Large language models may become widely available and similar to one another over time. What will remain unique is an organization’s own data—its customer feedback, research, product insights, and testing information. Clean, well-organized proprietary data will let companies generate insights that generalized systems can’t replicate.

Stambor: Do you think AI will remain a differentiator, or is it becoming table stakes?
Hanrahan: It’s quickly becoming table stakes. Leading organizations already use AI across purchasing, finance, product development, marketing, and customer engagement. Tasks that once required hours—gathering information, preparing documents, or analyzing product lines—can now be completed in seconds. Not adopting these tools creates an immediate disadvantage.

Stambor: An audience member asked how AI could contribute to healthier products. How do you see that playing out?
Hanrahan: Product development cycles are long and costly. When a launch represents a major investment, teams tend to reduce risk as they go, and over iterations, formulations often become higher in salt, sugar, or fat. If it becomes less expensive and less time-intensive to bring a product to market, teams can experiment more. Instead of betting everything on one item a year, they can try several ideas, including healthier ones. That creates more opportunities for nutritious options to reach shelves.

I’m also encouraged by the broader shift from focusing solely on nutrition panels to focusing on ingredients. Faster development cycles can support efforts to improve what goes into products.ft

This article is based on a live session at IFT FIRST. Responses have been edited for length and clarity.