Two years ago, the Ruiz Foods Board said they needed AI. "The board couldn’t actually define what AI was and needed me to explain the technology to them," said Michael Warter, SVP and CIO at Ruiz Foods Products, Inc., at the IFT FIRST event in 2024. Fast forward to Hannover Messe in 2026, and AI discussions are about actual applications and scalable data foundations, according to IT/OT Insider.

Management, plant managers and operations leaders are recognizing that context wrapped around data is the answer to scaling innovation and increasing margins. The first wave of context platform solutions is being applied to traditional data to produce advanced analytics, increased real-time visibility and to enhance overall equipment effectiveness (OEE) metrics.

The new technology solutions represent a step change in enabling real-time analytics for operators. "Data standardization is what allows OEE to shift from being a retrospective key performance indicator (KPI) to becoming a practical operating tool that teams actually use during the shift," says Shahzad Khan, Yokogawa Life Business Unit, global system consultant.

Incremental efficiency gains have management’s attention.

"When you examine the opportunities in yield and think about improvement, it's in millions of dollars," says Oliver Ganschar, head of digital product management and innovation at Kraft Heinz. "When there’s data in real time supporting a specific problem, it’s not hard to realize the possible value."

Explore More OEE

MontBlanc AI, a supplier of production monitoring software, released a case study on cooking oil processing and demonstrated how its modeling solution generated contextual data and anomaly alerts, preventing the degradation of over 120 tons of product. Operators eliminated manual interpretation of dozens of signals by modeling and applying contextual data, according to MontBlanc AI.

"The reality in today’s manufacturing world is that data doesn't have any context," says Aron Semle, chief technology officer at HighByte. "Changing the context of that data isn't that expensive, and companies need to start and iterate. The faster manufacturers contextualize data, expose it to AI and experiment with people on the factory floor, the better." HighByte’s Intelligence Hub’s industrial software directs modeled data to the cloud using a codeless interface.

An early adopter of HighByte's solution began modeling historical data and real-time information from its manufacturing execution system (MES) for a potato application. "They connected into our solution, added context to their historical data and fed it into the large language model (LLM) to create an internal chatbot," Semle says. "An operations expert began to chat with the chatbot, along with some guard rails, and it came back with interesting insights."

Food manufacturing leaders are seeking near-and-now solutions powered by data. "With existing and prospective clients, the mandate is to use AI, but it can’t be a multi-year process of deploying new systems," says Virag Vora, technical sales engineer at TwinThread.

"If unstructured data exists on paper or in disparate systems, the technology available today has come such a long way to merge all of that into a single context," Vora adds. "Companies can now leverage that data to train models and see measurable value quickly."

Advanced Analytics and OEE Solutions

The use of new digital tools to obtain accurate OEE metrics has been well-documented in recent years. Reported by FOOD ENGINEERING in February, Mark Bertrand, director, industry solutions at SmartSights, recently discussed how regression modeling and machine learning help find accurate "equipment culprits" for a specific packaging application.

For the last 30 years, the king of efficiency in manufacturing has been automation and control technology. Control Station developed the Overall Controller Effectiveness (OCE) metric for process automation. According to Control Station, "OCE is similar to OEE, but extends the concept to combine a PID controller’s availability, performance and quality into one score that can be rolled up from loop to unit to plant," says Bob Rice, vice president of engineering at Control Station, Inc. "In practical terms, OCE is a standardized and streamlined way to highlight which PID loops are helping (or hurting) stability, variability and setpoint tracking."

Setpoint monitoring via modeling introduces a new wrinkle for OEE formulas, with huge implications. "A company wanting to maximize an OEE model will dissect all the individual components, not only from a formula perspective of what contributes to OEE but then from an instrumentation or property perspective," Semle says. "There will be temperatures, pressures, line speeds, defect rates and all of these combined attributes are going to be correlated to that OEE."

New metric examples based on more granular data include energy per good unit, water per batch or yield losses tied to specific operating states like startup, changeover or clean-in-place (CIP).

This points to increased data visibility and real-time data for operators. "Once a (data standardization) foundation is in place, operators will finally be looking at the same version of the truth," Khan says. "That’s when OEE stops being reviewed at the end of the week and starts driving real-time decisions by operators, supervisors and engineers.

In the MontBlanc AI cooking oil application, early alerts enabled operators to intervene before degradation occurred, preventing oil from being downgraded to lower-value categories. "To no one’s surprise, the shift to increased granularity was the logical next step in process manufacturing environments, just as it was in discrete manufacturing applications," Rice says.

The TwinThread platform optimizes a range of process applications, including what batches, changeovers and startups look like. "Once companies have examples of all of those and add in historical data, then food producers can start training models for different scenarios," Vora adds.

In addition, models identify how certain operators or process engineers run a batch. "If you can codify that knowledge into rules that a model can extract, companies have a higher chance of success in reproducing the ideal changeover or the vertical startup. Context is still important," Vora says.

Micro-Stops and Success with AI

As context-aware data increases, operators will be able to leverage this information. Micro-stoppages are often missed in batch, filling and packaging applications. "Context-aware metrics quickly expose issues like micro-stoppages, recurring minor faults and gradual speed instability," Khan says. "Individually, these losses may not move availability much on paper, but collectively, they have a significant impact on throughput, waste and utility consumption."

Now, experienced operators – or humans-in-the-loop – can improve quality and reduce energy costs through context-aware data. "Operations in a real-time adjustment world can monitor ambient conditions and identify ways to improve the product, for instance, developing a new lot of flour with slightly more protein," Vora says. "This opens up an entire new world of micro-adjustments, but only if the data collection granularity supports it."

Food producers are aware of the success stories of machine learning applied to predictive maintenance. Mars recently piloted a predictive maintenance program for chocolate production across multiple plants using Datadog and Microsoft Azure’s IoT Edge platform.

"Data standardization is increasingly viewed as an operational requirement, not an IT initiative," Khan says. The low-hanging fruit of predictive maintenance has proved successful, which may be leading more food leaders and management to be sold on the need for data foundations and tools in 2026.