Kellanova taps Siemens AI to lift Pringles line 10% in Poland

The interesting thing about the Wall Street Journal's look inside Kellanova's Pringles plant in Kutno, Poland is what didn't get bought. No new fryers, no new stamping heads, no line rebuild. What went in instead was a network of sensors, lasers and cameras feeding temperature, humidity and protein readings into an edge-based Siemens AI model, plus a digital twin of the dough itself that keeps retuning the recipe as the raw potato materials shift. Siemens's claim, echoed on its own Pringles case page, is a 10% line performance lift, a 13% cut in waste and a 7% drop in energy use, none of it requiring new hardware.

Plant Director Ronny Matthijs frames the shift as moving from "the art of food making to the science of food making," which is a candid way of saying that a lot of what an experienced dough operator does is pattern-matching a moving target. Potatoes vary batch to batch. Humidity in Kutno on a wet Tuesday is not humidity in Kutno on a dry Friday. The value of feeding all of that into a live model is that the recipe stops being a fixed spec and starts being a control loop.

Why this matters beyond one snack line is the packaging. Siemens is folding this kind of work into its new Digital Twin Composer, unveiled at CES 2026, with PepsiCo already named as an early customer for the broader digital twin push. If the pitch travels — squeeze real percentage points out of an existing line by modelling the goo, not replacing the steel — capex plans across food and beverage for the next couple of years look different.

The honest caveat is that the headline numbers come from Siemens and its customer, not an independent audit, and vendor case studies tend to quote the good line. What the reporting doesn't give you is the cost of the sensors, edge compute and integration work needed to hit those figures, or how transferable the gains are to plants without Kutno's data plumbing. If the economics hold up, the near-term winners are Siemens, Kellanova and any operator sitting on an aging line where the fastest capacity is the capacity you already own.