Delivering Context to Food Safety Data at the Machine Level

Data-first architectures have changed how plants approach operations at both greenfield and legacy plants. The next phase is adding context to food safety data at the equipment level or in middleware.

“Taking months of data down to a single day,” said Keven Wang, Co-Founder and CEO of UnitX, at Automate 2026, held June 22-25 in Chicago. UnitX supplies an AI vision inspection platform and, during the panel discussion, Wang stated that generative AI is a key tool for accelerating the detection of potential defects in products. The upshot, from Wang, is that generative AI identifies defects quickly and provides a library of them in less time than traditional pilot programs.

UnitX technology is suited for high-margin industries, such as batteries and automotive parts, not necessarily food and beverage. However, midsize processors and food majors are capturing food safety data at the machine level, and the next phase is adding context via equipment software or middleware.

Food safety transition

Multilayer security processes protect machines and analytics, and limit the ability of hackers to move into data software or middleware at plants.Blentech

Industrial networking and data sharing are winning the day at greenfield plants. “Twenty years ago, it was solely about getting the equipment up and running,” said Bob Rice, VP of Engineering at Control Station, Inc., in a 2025 ProFood World article. “Now, many big projects are coming in with (operations) standards where you have to reach a certain production level and start applying analytics well before the first project."

Industrial networking and automation advances during the last 15 years have lowered the barriers to entry for plants and machine suppliers. “Prices have come down to the level where robust adaptive recipe management with quality built in can be done either by smaller processors or lines that have more variability,” says Daniel Voit, CEO and CRO of Blentech.

“The industry has moved from experimentation into operation, and this seems to be the expectation for new equipment,” adds Voit. Blentech supplies mixing and blending equipment, along with a turnkey machine-level analytics solution.

With data-sharing expectations in place, food producers want more context and visibility into food safety data. “Using OPC UA connectivity and a data collection system, multiple machines can be networked to provide a single view of inspection performance,” says Eric Garr, Regional Sales Manager at Fortress Technology.

Fortress Technology’s data collection system, Contact 4.0, provides processors with a full suite of compliance reports for metal detector performance verifications, tracks overall equipment effectiveness (OEE), and feeds performance insights directly to Quality Assurance (QA) and technical teams.

Fortress Technology has more than 50 metal detectors at a dry rice facility with more than 100 long- and medium-grain rice brands. The detectors are used across the plant, including processing, packaging, and lab operations. Legacy machines are part of the solution and have been retrofitted to transmit food safety data via the company’s software.

“Our customers can use a data monitoring tool, which sits on the main machine network and collects data on the fly,” adds Garr. “It’s a web-based tool and allows operators to use a web browser to monitor inspection points and also be used as a notification system.”

Food producers are buying equipment, but this example also shows that legacy machines can thrive. “An entire component of our business is dedicated towards upgrading and retrofitting legacy equipment supplied by external batch processors and us,” adds Voit.

Automation advances propel food safety data

While the food industry embraces AI and struggles to implement AI strategies, the transition to fully automated systems continues. Quality is king in today’s production environment, and automated processes elevate operators' roles while ensuring compliance with food safety standards.

With food producers searching for higher throughput, data collection at the machine level is redistributing the workforce toward production.Fortress Technology“We have recipe systems built to integrate quality steps so you can't progress on with the recipe and can't discharge the product unless it meets the criteria,” says Blentech’s Voit. “Unless you have the appropriate login level, you can't get it out of the machine, and that's a way of guaranteeing accurate data.” One example includes barcode scanning of ingredients for batch production.

Increasing throughput while maintaining the current workforce is a dominant mantra in operations. “That’s been a huge advancement in metal detection is eliminating human error with performance verifications, especially in smaller operations,” adds Garr. “Quite often the operator responsible for production also did quality checks on the machines, so production is benefiting.”

Going forward, midsize processors and food majors are deciding how to label data and add context to support real-time decision-making in processing and packaging. Blentech uses a cloud-based analytics platform called ARTIS to provide customers with trend visualization, Quality Assurance (QA) compliance reports, and remote recipe control.

The platform uses Inductive Automation Ignition technology, MQTT communication protocols, and Amazon Web Services (AWS).

“A lot of smaller processors or customers that don’t have AI modeling capability use our platform for batch processing data,” says Voit. “The key is data context and whether or not the information streaming into the systems can be consumed by the AI tool.”

Moreover, data labeling for factory floor operations doesn’t require a data scientist on staff. In-house experts, along with OEMs, can apply labeling to operations data. “Did you label the temperature ‘TT01’ or temperature probe one, two, or three?” says Voit. “If the program, software and context engine is structured to understand the data properly, you can make the data far more valuable for AI.”

AI approaches for food majors

Chick-fil-A and Cargill gave updates on their AI programs at the Food Safety Summit in May. Tracking suppliers is a major pain point for these food players, and the move from Excel-based processes to digital tools is underway.

“We are targeting our greatest risks using quantitative microbial risk assessments (QMRA),” said Steven A. Lyon, Director of Food Safety–Field Operations at Chick-fil-A. “Modeling our highest risk is our strategy and focus with AI at the moment.”

“It comes down to managing risk across supply chains when huge swaths of data are available and they're not connected.” — Sean Leighton, Global Vice President of Food Safety, Quality, and Regulatory at Cargill

The restaurant company models at the supply chain level and tracks salmonella at critical control points in the supply chain. Chick-fil-A uses the Ancera AI platform to monitor salmonella risk in real time, focusing on predictive analytics rather than just testing, according to reports from the event. Limited details were revealed, but the process moves siloed data to a centralized database and applies AI models to help decision-makers.

“AI is great for awareness, but human beings must act on it,” said Lyon. “AI is here to make humans better at risk-based decision-making.”

The overarching theme at the conference was siloed data and labeling efforts. “It comes down to managing risk across supply chains when huge swaths of data are available and they're not connected,” said Sean Leighton, Global Vice President of Food Safety, Quality and Regulatory Affairs at Cargill.

Leighton described how disconnected data led to traceability issues in the locust bean gum recall. “Locust bean gum was discovered to have trace amounts—less than one part per billion—of ethylene oxide in certain lots of turkey, and this recall ended up being the most costly of my 25-year career.”

In other areas of the food industry, Danone is discussing agentic AI. The dairy major’s work with Microsoft has shifted from training and productivity tools into autonomous agents, while its DanSkills program remains focused on upskilling employees for AI-enabled work.

Data context remains the current challenge, but the next phase is vision and imaging. “We are building systems that allow timestamped image importation for visual analysis for aggregation of data,” says Voit. “Take pictures of the picked product with a mobile device, bring it into the dataset, attach it when it was taken, and the data gets correlated with the broader operational data that is already contextualized.”

“Integrating vision into the whole footprint will help track packaging as retail standards evolve,” says Garr. “The future will have serialized food products on retail store shelves and build out traceability, along with potential defects.”

Although FSMA 204’s compliance date has been pushed to July 20, 2028, food safety data advancements continue across the industry.