Beyond the AI Hype

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

As AI adoption grows, expectations for what these systems can do are not always aligned with reality. AI can identify patterns and highlight potential risk, but it cannot confirm contamination or replace human judgment. In this Q&A, Willette Crawford, owner and principal of Katalyst Consulting LLC, explains where the technology adds value and why validation and oversight are critical.

Where do you see AI generating measurable value in food plants?

AI is generating the most measurable value in areas where data is continuous, structured, and directly tied to decisions such as environmental monitoring, sanitation verification, and process control.

The real impact isn’t automation, it’s prioritization. AI helps teams recognize meaningful signals earlier and act before a loss of control occurs. In that way, it strengthens preventive controls by improving timing and consistency, rather than replacing the system itself.

Where does the hype around AI exceed the science?

The hype exceeds the science when AI is positioned as a detection tool rather than what it actually is: a pattern recognition system built on imperfect data. AI doesn’t detect pathogens; it interprets signals that may correlate with risk. Without strong data quality, context, and validation, it can create a false sense of precision.

So, the danger isn’t the technology itself—it’s overconfidence in what it can actually prove. The effectiveness of AI depends heavily on the quality of the data and the context in which the system is built.

The real impact isn’t automation—it’s prioritization, helping teams act before a loss of control occurs.

Willette Crawford

Katalyst Consulting

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

AI doesn’t remove or shift accountability. If anything, it makes governance more critical in food safety systems. Companies are still responsible for the decisions they make, whether those decisions are informed by AI or not. If a system fails, the issue isn’t just the model error, it’s whether the system was appropriately designed and validated, and whether proper oversight was in place. Responsibility ultimately sits with the organization, not the algorithm.

How should companies validate an algorithm that is making or influencing food safety decisions?

Validation has to go beyond a one-time exercise. It should mirror process validation, but with continuous performance monitoring. That means testing the model against known outcomes, verifying performance under real-world variability, and reassessing as conditions change.

Models drift, and operations evolve. So, validation needs to be ongoing, using real-world experience to continually assess whether the system is performing as expected.

Where should human judgment remain central, even as AI systems become more sophisticated?

Human judgment remains essential in risk interpretation, escalation decisions, and contextual understanding. AI can identify patterns, but it cannot fully interpret operational nuance or accountability. The most effective systems are “human in the loop,” where AI sharpens expertise rather than replaces it. In food safety, judgment and responsibility will always be human functions.

How should companies communicate algorithm-driven decisions to regulators?

Companies should communicate AI systems based on their intended use. They shouldn’t present AI as a black box or a magic system. Instead, they need to clearly explain the intended purpose, the inputs and outputs used to design the model, the decision thresholds, and the human oversight and validation processes behind it. That level of transparency gives regulators the information they need to feel comfortable with how the system is being used.

What transparency requirements might regulators expect from AI systems?

Regulators will expect many of the same fundamentals they already require but applied to AI systems. That includes traceability of data sources, documentation of the model’s purpose, evidence of validation, and a clear explanation of how outputs are used. They will also expect records of overrides and changes, along with clarity about who remains accountable for decisions. At the core, regulators will want to see that companies understand how the system works, how it is governed, and how it fits into the overall food safety program.

This interview has been edited for clarity and brevity.