AI and digital ontologies like FoodOn aim to fix food R&D’s data problem

The challenge of fragmented food data

Food research and development (R&D) faces growing complexity due to fragmented data across languages, systems, and regulatory frameworks. According to YourStory, this fragmentation complicates product development, increases safety risks, and slows compliance with market-specific requirements.

The World Health Organization reported that unsafe food caused 866 million illnesses and 1.5 million deaths globally in 2021. Chemical hazards, such as inorganic arsenic and lead, were responsible for 73% of these deaths, with arsenic alone accounting for 42% and lead for 31%. The economic impact was equally stark, with $310 billion in lost productivity due to foodborne diseases.

These risks highlight the need for connected, traceable data systems to identify hazards early and reduce costly rework or recalls.

How digital ontologies like FoodOn help

Digital ontologies, such as FoodOn, aim to solve data fragmentation by creating a shared vocabulary for food-related information. FoodOn, an open ontology, now includes tens of thousands of classes covering food products, their sources, and production processes.

By standardizing terminology across languages and systems, ontologies enable AI to interpret and connect siloed data more effectively. This structured approach reduces errors and improves the reliability of AI-driven insights.

AI’s role and its limitations

Generative AI has the potential to streamline food R&D by analyzing supplier documents, scanning scientific literature, and suggesting alternative ingredients. However, its effectiveness is limited by fragmented or unstructured data, which can lead to unreliable or unverifiable outputs.

For example, YourStory notes that AI struggles to predict sensory outcomes, such as taste or texture, or assess how ingredients behave in real-world conditions like humidity. Human expertise remains essential for validating AI recommendations, particularly in safety and regulatory contexts.

The future of data-driven food R&D

The integration of AI and structured data systems could transform food R&D by embedding safety and compliance into the development process from the outset. This shift may favor companies with organized,machine-readable data over those relying on more powerful—but less grounded—AI models.

Early identification of risks, such as contaminants or regulatory non-compliance, could reduce delays and prevent costly recalls. However, the need for human oversight ensures that AI remains a tool rather than a replacement for expertise.

What this means

Lazyfounder analysis — our interpretation, not reported fact.

For founders and operators in the food and agriculture sectors, this shift toward structured data and AI-driven R&D presents both an opportunity and a challenge. Startups developing plant-based proteins, functional ingredients, or wellness-focused products are often constrained by the time and cost of ensuring safety and compliance across markets. AI and ontologies like FoodOn could accelerate development cycles by providing a unified framework for interpreting regulations, supplier data, and scientific literature.

However, the reliance on human validation means that companies cannot solely depend on AI for critical decisions. Teams must invest in organizing their data in ways that both people and machines can interpret—a competitive advantage in an industry where fragmented information remains a bottleneck. Operators should also consider how ontologies might integrate with existing workflows, particularly if they operate in multiple markets with varying compliance requirements.

The economic stakes are high. The $310 billion lost annually to foodborne diseases underscores the cost of inefficiency, while the dominance of chemical hazards like arsenic and lead in rice highlights the need for proactive risk management. For startups, this could mean rethinking how data is collected, stored, and shared—not just as a compliance requirement, but as a strategic asset.

Related on Lazyfounder

• World Health Organization
• More ai news