Optimization, Not Automation

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

The impact of AI in food safety is showing up in practical ways. It is helping teams use data more effectively, identify risk sooner, and respond before problems escalate. In this Q&A, Matt Henderson, vice president of technical services at Land O’Frost, explains where AI is already delivering value inside food plants and how it builds on approaches companies already use.

How would you describe the current role of AI in food safety systems—are we still in the experimental phase or is it becoming operational?

I think it is becoming operational, but it’s a broad spectrum when you talk about AI. On one end, you have predictive analytics tools. Those are still being developed, and a lot of companies—including us—are testing them. I would say those are moving beyond experimental and into implementation mode, but they’re probably still a year or two away from being fully realized at scale.

On the other hand, there are already very practical applications that we’re using every day. For example, generating training materials, drafting SOPs, or quickly developing content when a topic comes up. A food safety manager can put a prompt into a

n AI tool and generate materials that would have taken much longer before. That’s a real operational improvement.

So, I’d say we’re already seeing operational gains in some areas, while the bigger-picture predictive modeling capabilities are still developing.

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

I think the biggest opportunity is in prediction. The industry has long aspired to move from preventive to predictive—implementing controls before the likelihood of failure presents itself. AI makes that more attainable because of the sheer amount of data analysis required. A lot of companies simply don’t have the tools or resources to analyze that level of data on their own. AI can process those datasets and identify patterns in a way that enables that predictive stage.

In practical terms, I think about something like Listeria control. We’ve historically used approaches like post-rinse sampling or other early warning indicators to identify potential risks before they reach product contact surfaces. If AI is telling me that conditions are trending in a risky direction, that’s really just a more advanced version of what we’ve already been doing. Food safety isn’t compromised yet, but we’re getting information that it might be headed that way.

[With AI], instead of reacting after a problem occurs, we can take action earlier, before intervention becomes necessary. That’s where the measurable value comes in.

How should companies respond to AI-generated early warning signals?

In my mind, it’s the same response framework we’ve always used with early warning indicators. If food safety is not yet compromised, but the data suggests it could be trending that way, then you don’t necessarily stop the line. But you do take action in advance of that being required. That’s the key difference—AI allows you to act earlier. It’s an enhancement of existing systems, not a completely new way of thinking about response.

Predictive analytics is really about optimization—I can bias my resources toward where failure is more likely.

Matt Henderson

Land O’Frost

How does predictive analytics change the way companies approach risk?

Predictive analytics is really about optimization. The risk is still present. I’m still doing my risk assessment and validation exercises, but what changes is how I allocate resources. For example, instead of sampling all locations at the same frequency, analytics may show that risk is more likely to occur in a subset of locations. That allows me to bias my resources toward those higher-risk areas. So, I’m still achieving the same food safety outcome, but I’m doing it more efficiently.

That’s really what predictive means in practice—it’s achieving the outcomes we’re already capable of in the preventive stage but doing it in a way that creates business upside as well. If I can maintain the same level of food safety while optimizing resources, that’s a win for the organization and for the consumer.

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

At the end of the day, it still relies on an individual. AI is a tool. It can support decision-making, but it doesn’t replace it. For example, if I’m using predictive analytics to optimize sampling, the tool might give me insights about where to focus. But I still have a person making the decision that it is in a food-safe condition before the product is released from the facility.

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

Human judgment remains essential in system design, validation, and final decision-making. Using a HACCP framework, we need to understand potential failure modes. If one of those failure modes is that AI could misinterpret something, then we need an intervention—typically a human verification step. A person is still responsible for validating how the system is used, whether that’s validating a piece of equipment or reviewing a critical control point before product release.

Even beyond that, context matters. If I had a very strong data analyst who understood the numbers but didn’t understand food safety or how the plant operates, I wouldn’t have that person making release decisions. The person making that decision needs to understand the system and be able to interpret everything in context. AI can provide insights, but it can’t replace that level of understanding.

How do you view concerns about overreliance on AI or errors in AI-generated outputs?

I look at it from a management perspective. If someone uses a chatbot to generate work and doesn’t review it, and the output is poor, that’s no different than someone rushing to complete work at the last minute and turning in something flawed. Either way, it’s still poor work.

The same applies to food safety. If you rely on a tool without properly reviewing or validating the output, that’s still your responsibility. The tool doesn’t change accountability.

Is there anything else food safety professionals should be thinking about as AI becomes more integrated?

The scenario you described—where AI identifies a developing trend—is exactly what companies will be facing more often. We’re going to have situations where data tells us that risk may be increasing, even though food safety hasn’t been compromised yet. In those cases, I would treat AI as an early warning indicator—similar to the ones we’ve already built into our systems, just more advanced.

The response should be the same: Recognize that food safety is still intact but take action in advance based on the signals you’re receiving. That’s where AI fits best—helping us act sooner, not replacing the systems or decisions we already rely on.

This interview has been edited for clarity and brevity.