Training for Tomorrow: How to Leverage Artificial Intelligence in Food Safety
Article By Jacqueline Mitchell, Editor, QA Published July 22, 2025
Article Source: https://www.qualityassurancemag.com/article/how-to-leverage-artificial-intelligence-in-food-safety/?utm_campaign=Quality+Assurance+%26+Food+Safety+News&utm_source=08%2f05%2f2025+-+%5bQA+News%5d+Reagan-Udall+Foundation+Releases+Produce+Safety+Roadmap&utm_medium=email&utm_term=https%3a%2f%2fwww.qualityassurancemag.com%2fArticle%2fhow-to-leverage-artificial-intelligence-in-food-safety&utm_content=1047096&isid=B5E31F
In the age of AI, food safety and quality assurance pros are the new data detectives. Purni Wickramasinghe of Chick-fil-A reveals how smart training and sharper insights on artificial intelligence in food safety can turn complex data into life-saving decisions.
The key to a safer food supply may lie in the data we already have — and the people trained to unlock its potential, with the help of artificial intelligence (AI).
Increasingly, artificial intelligence in food safety is becoming a foundational tool to enhance human capabilities and drive smarter, faster decisions across the industry.
“The goal of AI should not be to replace human thinking,” said Purni Wickramasinghe, food safety and restaurant solutions lead at Chick-fil-A, during a presentation at the 2025 Food Safety Summit in Rosemont, Ill., in May. “The goal of AI should be to enhance human thinking — to empower humans to think through complex data sets so that we can make informed, accurate and fast decisions to, in this case, ensure food safety.”
Wickramasinghe delivered a call to action for the food safety and quality industry: For AI to meaningfully transform food safety and quality assurance, professionals must invest in the education and skills necessary to interpret and guide intelligent systems.
COMPLEXITY DEMANDS MORE.
Food safety, while grounded in basic principles, operates in an increasingly complex and high-stakes environment.
For example, Wickramasinghe said, if she asked her husband (a data engineer), about food safety, he would say, “It’s pretty simple. You cook your chicken to 165 degrees [Fahrenheit], hold your cold things to 41 degrees [Fahrenheit] and below. You don’t leave food out at room temperature for too long; put it in the fridge. That’s about it.”
“But those of us who live in food safety, who get to work in food safety, know it’s a complex system,” said Wickramasinghe. “The complexities keep coming at us.”
That complexity stems from several fronts: human variability in operations, increasingly globalized supply chains, nuanced regulatory interpretations across jurisdictions and the unpredictable behavior of pathogens under varying environmental conditions.
“We are leaning on 16- and 17-year-old team members to execute complex processes inside the restaurant,” Wickramasinghe said. “We are asking them to do a lot.”
Wickramasinghe likened Chick-fil-A’s 3,000 restaurants to 3,000 mini laboratories, each with its own set of variables. There are manual checks, global supply issues and equipment inconsistencies — then add foodborne pathogens like Listeria and E. coli, each with their own growth patterns and infectious doses.
These challenges produce vast amounts of data. Yet, Wickramasinghe noted, most of that data lives in silos: on paper logs, in disconnected software systems or as untapped sensor readings.
“Food safety is a very complex system that generates a lot of complex data,” said Wickramasinghe. “How do we ensure that we look at all this data that’s interconnected to influence a decision? We need to leverage something that’s more than just you and me. We need to leverage a tool, an infrastructure, and that’s where AI comes into play.”
TRAINING FOR THE AI ERA.
Central to Wickramasinghe’s message was the idea that AI implementation is only as good as the humans who guide it. That requires specialized training — some of which is well outside the traditional toolkit of a food safety manager.
To succeed, companies need to build teams that combine traditional food safety knowledge with skills in data science, engineering and analytics, she said.
“You have food safety knowledge, whether it’s in the public health, microbiology or operation space,” said Wickramasinghe. “It’s vital. So keep that. But what I would encourage you to add on is to take those online courses, the elective courses around data science.”
Food safety and quality assurance professionals, she suggested, should pursue training in several key areas:
Data Science Fundamentals. “Not just analytics, like looking at big data sets in Excel to be able to get an average and a standard deviation,” she clarified. “I mean really try to understand data science principles. What are the different kinds of models that are out there?”
Coding and System Languages. “You need to learn SQL to query your data … and understand Python or whatever your preferred system language is for your organization,” she said. “If you have the ability to learn it, learn it.” Fun fact: ChatGPT can teach you how to code in Python. “You don’t need to look very far,” said Wickramasinghe. “You don’t have to pay for a course. You can leverage AI that’s available to you to learn so that you can do really well at AI for your organization.”
Communication and Storytelling. Perhaps just as critical as coding, Wickramasinghe emphasized the ability to tell an impactful story with data. “The way you take away that fear factor associated with AI, or just inherent lack of interest in adopting it, is to tell a really good story and lean into the narrative,” she said. “Outside of the data side of it, the engineering side of it and the microbiology side of it, get really good at communicating to the right audience in the right way.”
MODELS BEHIND THE MAGIC.
While most audiences associate AI with tools like ChatGPT, Wickramasinghe walked attendees through a more nuanced understanding of AI’s practical applications in food safety — from simple regression models to complex machine learning systems.
“ChatGPT is one type of AI model,” she said. “That is a large language model. There are many other types of models. For those of us who took statistics classes, a regression model is still an AI model. It’s a very primitive one, but it’s an AI one. You can use different types of models to understand and learn from your data. That eventually gets you to a place where you are AI- and data-informed in your decision-making, whether it’s risk assessment or prescriptive analytics.”
Wickramasinghe highlighted the following models and provided some food safety applications:
Predictive Model. Use research data to predict microbial growth based on time and temperature. For example, “If I hold my chicken at 45 degrees [Fahrenheit] for six hours, this is what the Salmonella growth looks like,” she said.
Machine Learning Model. Use a decision tree model to predict product temperature by analyzing equipment data, including how often doors get opened, and ambient temperature. “How long is the door held open for? You can factor these variables in to be able to predict your product temperature and the impact on your product,” said Wickramasinghe.
Simulations. Use a Monte Carlo simulation (a technique that uses random sampling to model the probability of different outcomes in a process that involves uncertainty) to model how norovirus might spread from a single infected team member to multiple surfaces and food items — and how various handwashing protocols might reduce that risk.
Real-Time Monitoring. Create systems that track environmental conditions continuously to move from reactive to proactive food safety management.
Decision Support Systems. Some companies build custom AI tools similar to ChatGPT trained on internal company documentation. Instead of Googling the correct cooking temperature, your team can ask your system directly, Wickramasinghe said. “This is your own version of ChatGPT, however you want to name it,” she said.
FROM INSIGHT TO IMPACT.
As Wickramasinghe pointed out, the power of artificial intelligence in food safety holds little meaning without human action.
“You get to interpret the results,” she said. “The model can put out any output. You as a scientist, as a businessperson, get to analyze the outcome. You get to gather the insights. You get to decide the risk threshold. We unfortunately don’t exist in a zero-risk world, so we need to have a risk threshold that we are comfortable with. You get to define it, and you get to identify based on the model, or multiple models, where are you? Is this new process you’re introducing to your restaurant or business going to eat into your risk threshold, or is it going to expand your risk threshold? You get to make that decision.”
Human oversight is necessary not only for decision-making but also for finetuning model accuracy, determining contextual relevance and ensuring the model evolves with changing business needs, said Wickramasinghe.
“Humans add meaning to AI and data,” she said. “At the end of the day, we can have all the complex models and all the good data, but if humans are not there to tune, interpret, decide, restructure and act on our leads and provide context, that doesn’t mean anything. Humans are an integral part of this equation.”
BUILDING A COLLABORATIVE FUTURE.
Wickramasinghe addressed common fears about AI around job displacement, urging the industry to move past the notion of AI as a threat.
“There’s going to be natural cultural resistance,” she said. “There’s this sentiment that AI is here to replace my job. People are going to feel some type of way about AI. Let’s expect the resistance, embrace it and then talk them through it. Bring them along the journey that AI and data doesn’t mean anything without you and your interpretation and your contribution.”
TRUST AND THE ROAD AHEAD.
As with any high-stakes system, food safety AI must be implemented with care. Wickramasinghe noted challenges including data reliability, legal considerations and infrastructure maintenance.
“There’s no room for error,” she said. “We’re in a high-stakes space. Food safety is life or death in some cases. You don’t want wrong AI predictions.”
To navigate these risks, she encouraged building trust through transparency — openly communicating model limitations, how predictions are made and what actions follow.
“Then you get to educating,” she said. “You don’t want to just educate your data team. You want to educate your entire team: the decision makers, the directors, the VPs, the execs. ... If we do this, and we do it really well, we get to a point where everybody trusts in AI, and we can actually turn our insights into impact.”
A CALL TO ACTION.
Wickramasinghe left FSQA pros with a challenge for the future as they learn to implement artificial intelligence in food safety: “Let’s build a safer future together. AI gives us tools. It’s an infrastructure. But it takes people to make it work.”
She pointed to a vision of a fully integrated system where food safety professionals combine microbiological knowledge, data literacy and communication skills to drive informed decisions.
“You’re bringing data, you’re bringing AI, you’re telling a story, and you’re bringing you along that journey to influence food safety for the safety of all humans,” she said.
For the food and beverage industry, it’s a future that’s already here. What remains is whether its professionals are ready — and trained — to lead it.
Editor’s Note: Information for this article was taken from a presentation, “Leveraging AI for a Safer Today and Tomorrow,” at the Food Safety Summit in May 2025.