What Comes After AI Insights?

As artificial intelligence (AI) makes information easier to find, synthesize, and act on, the harder question for food innovators may be what comes next: Which problems are worth solving? Which signals point to real opportunities? And where does human expertise still make the difference?

Those questions framed a Scientific & Technical Forum at the IFT FIRST Annual Event and Expo that brought together perspectives from consumer health, food R&D, digital product development, and sensory science. The panel included Pelin Wood Thorogood, co-founder and executive chair of Radicle Science; Jay Gilbert, director of CoDeveloper and Digital Products at IFT; Anshul Dubey, R&D transformation and digital capabilities lead at PepsiCo; and John Ennis, CEO of Aigora.

Gilbert put the challenge plainly during an opening lightning round. As information becomes more abundant, he said, “The thing that becomes scarce is really the thing that makes us all human. It’s our curiosity, creativity, the ability for us to connect with each other.”

The thing that becomes scarce is really the thing that makes us all human. It’s our curiosity, creativity, the ability for us to connect with each other.

Jay Gilbert

IFT

More Data, Different Questions

One major shift is the amount and variety of data researchers can now analyze. Wood Thorogood described work that combines physical and digital biomarkers, wearable data, behavior, diet, and other real-world inputs. Rather than stripping away variability as noise, she said researchers can increasingly examine that noise as context—information that may help explain why people respond differently.

That opens the door to a different kind of scientific inquiry. AI and machine learning can search for patterns researchers did not necessarily know to ask about, potentially generating new hypotheses around personalized nutrition and health. Wood Thorogood cited a recently launched food-as-medicine study examining multiple prebiotic interventions alongside multidimensional health data as one example of how those methods are expanding.

The information about what didn’t work is actually extremely valuable.

John Ennis

Aigora

Don't Lose the Dead Ends

The panel also pointed to a less obvious opportunity: using AI to make better use of what organizations already know.

Food companies have accumulated years of R&D reports, consumer studies, formulation records, and other internal information. AI can help organize and search that material at a scale that would be difficult to manage manually. But Ennis cautioned that corporate knowledge bases often preserve the successes while losing the experiments that failed.

“The information about what didn’t work is actually extremely valuable,” Ennis said. An AI system searching for a solution needs to know which paths have already been explored and why they failed, he explained, or it may spend time rediscovering dead ends that an experienced scientist could have ruled out immediately.

Gilbert extended that concern to tacit knowledge—the practical expertise that can disappear when employees retire or leave. Capturing what experienced people know, including what they tried and rejected, could make institutional knowledge more accessible to the next generation of scientists and to the AI systems supporting them.

Where AI Works Now—and Where it Doesn't

Across the innovation pipeline, Dubey identified several areas where AI is already delivering value, as well as gaps that remain.

Delivering value now:

• Synthesizing consumer, market, and scientific information
• Creating and testing concepts more quickly
• Supporting compliance checks and regulatory validation
• Finding quality signals in manufacturing data

Still developing:

• Identifying weak signals that may become consumer needs three to five years from now
• Formulation and discovery involving new ingredients or scientific mechanisms
• Capturing tacit organizational knowledge consistently
• Producing reproducible scientific insights across every use case

Dubey drew an important distinction between insights and foresight. Companies are getting faster at understanding what is happening now, he said, but identifying the faint signals that could become meaningful trends several years from today remains much harder. Scientific discovery and formulation are also advancing, but many of those applications are still emerging.

It takes a human to decide what’s worth working on.

John Ennis

Aigora

Treat the Answer as a Starting Point

The panel's message was not to wait for perfect technology. It was to be clear about what a tool can and cannot do—and to validate what it produces.

Ennis advised users to treat AI-generated work as a draft, interrogate the basis for its claims, and return to the underlying sources. The validation process, he noted, can itself become a way to learn. Panelists also cautioned against expecting one platform to solve every problem; organizations should begin with the problem they are trying to solve, then determine whether they have the data and expertise to address it internally or need a specialized outside tool.

That approach keeps the technology in its proper role. AI can accelerate research, narrow options, surface patterns, and reduce some of the time required to move an idea forward. But the panel repeatedly returned to the decisions that still belong to people.

“It takes a human to decide what’s worth working on,” Ennis said.

Explore IFT FIRST Coverage

Discover the latest reporting, interviews, and expert insights.

Read Now