Getting the Data Right

This interview is part of our extended Q&A series exploring AI and food safety.

For AI systems to work in food safety, the challenge often is not the technology itself. It is the data behind it. Companies are still working to organize, standardize, and validate the information these systems rely on. In this Q&A, Hal King, managing partner of Active Food Safety and former director of food safety at Chick-fil-A, explains why data structure matters and how organizations are beginning to integrate AI into existing business systems.

How would you describe the current role of artificial intelligence in food safety systems? Are we still in the experimental phase, or is it already becoming operational in some parts of the industry?

Food businesses are developing better systems to enable the collection and formatting of data for harmonization, which is essential for using AI effectively. Proper data format is key to enabling AI-driven analytics and reporting, including generating recommended actions that can be trusted without introducing AI-related errors.

Some companies are already using trained AI systems as a “source of truth” across their enterprise. For example, organizations are training AI using their operational SOPs and expected corrective actions so that employees at any level can prompt the system and receive guidance on proper actions—whether that involves equipment maintenance, correct operational procedures, or environmental monitoring responses.

Proper data format is key—without it, you can’t trust what AI is telling you.

Hal King

Active Food Safety

The technology is also evolving rapidly at the enterprise level. Many food safety–specific needs—such as tracking KPIs or initiating corrective actions—are beginning to be embedded within broader business functions like operations or supply chain, where responsibility for execution and verification already resides.

What mistakes do companies make when adopting AI tools?

One of the biggest mistakes is not having data in a format that is usable. If the data are not structured properly, it becomes difficult to trust the recommendations generated by the AI system.

Another common issue is implementing AI without what I call “integrity gates.” These are validation steps—either human oversight or secondary automated systems—that verify outputs before action is taken. Without these safeguards, companies often end up doing more work to manage the system rather than gaining efficiency.

Food safety systems have historically relied on verification and testing. How does predictive analytics change the way companies approach risk?

Right now, predictive analytics is helping businesses determine what should be audited or tested, which can reduce the cost of audits and testing. Looking ahead, these systems will likely evolve to include prompts that verify system elements in real time—such as visual observations through cameras or personnel—focused specifically on areas where hazards are no longer controlled. That represents a shift toward more targeted, risk-based verification rather than broad, uniform testing approaches.

If an AI system fails to detect a contamination risk—or produces a false signal—how should responsibility be handled?

At this point, I’m not aware of any businesses that are allowing AI outputs to inform final food safety decisions on their own—and they shouldn’t. Human oversight remains essential, particularly when decisions could impact the safety of food.

What practical advice or watchouts can you share with food company decision-makers about applying AI in food safety systems today?

First, ensure that your data—such as KPI data—are in a format that works effectively with the AI system you are using, and validate that data regularly.

Second, implement integrity gates to verify outputs so you can gain real value from AI without introducing risk. These validation steps are critical to maintaining confidence in the system.

Third, no decisions that could affect the safety of food—whether in foodservice or manufacturing—should be made by an AI system alone.

Finally, data security is a major concern. Companies need to ensure that AI systems do not publicly share or expose confidential information, whether through unintended outputs or cybersecurity vulnerabilities. For that reason, many organizations should prioritize internal, enterprise-controlled AI systems.

This interview has been edited for clarity and brevity.