AI’s real pet food value is in closing the formulation gap

Bottom line

Pet food manufacturers are getting a clearer message on AI: the technology is most useful when it connects formulation, production, and quality data, not when it sits in a standalone prediction tool. Recent industry coverage tied to Petfood Forum 2026 highlighted that gap, with BESTMIX’s Carmen Sook arguing that quality and efficiency still depend too heavily on operator judgment, delayed QC feedback, and siloed systems. In one Alltech case study, a moisture prediction model built from four years of plant data improved in-tolerance performance from about 60% to 87%, cut startup stabilization time by 43%, and was projected to save roughly $567,000 annually across two extrusion lines. (petfoodindustry.com)

Why it matters: For veterinary professionals, this is manufacturing news with nutrition implications. More consistent extrusion and moisture control can affect kibble texture, coating uptake, palatability, and batch-to-batch uniformity, all of which shape whether a finished diet performs as intended for pet parents and patients. The broader takeaway is that AI won’t compensate for weak process control or bad data; its value comes when manufacturers use plant-floor information to tighten formulations, reduce rework, and catch deviations earlier. That matters in a category where formulation errors are often driven by data and process failures upstream, not just by nutrient targets on paper. (petfoodindustry.com)

What to watch: Expect more pet food companies to pilot closed-loop AI tools in extrusion and QC first, then expand into formulation support as data collection becomes more automated. (petfoodindustry.com)

AI may be one of the loudest themes in pet food manufacturing right now, but the emerging industry consensus is more restrained: prediction alone doesn’t fix production. The real opportunity is in closing the loop between formulation, line settings, and quality outcomes so each production run improves the next. That’s the premise behind recent reporting from GlobalPETS and related coverage from PetfoodIndustry, which framed AI less as a magic layer and more as infrastructure for continuous improvement. (petfoodindustry.com)

That framing reflects a longstanding problem in pet food plants. Even with automation and formulation software, operators often still make line adjustments based on experience, while quality data arrives with a delay and formulation teams work from separate systems. According to Carmen Sook of BESTMIX, those disconnects create three persistent gaps: shift-to-shift variation in operator judgment, about an hour of lag before QC results are visible, and no true feedback loop linking formulation, production settings, and QC results. (petfoodindustry.com)

The strongest concrete example so far comes from an Alltech deployment described at Petfood Forum 2026. Using four years of production data, including roughly 27 logged parameters per hour alongside recipe and ingredient data, BESTMIX built a model to predict finished moisture before product reached the dryer. Tested against historical runs, the model improved moisture accuracy from about 60% to 87%, reduced startup stabilization time by 43%, and generated projected annual savings of about $567,000 across two extrusion lines. Most of that value came not from flashy automation, but from less rework, tighter formulations, and faster startups. Alltech has since linked its manufacturing execution system with BESTMIX QC to move from hourly manual logging to automated data capture every six minutes. (petfoodindustry.com)

That fits with a broader shift in industry messaging. A recent BESTMIX-sponsored article argued that many AI projects underperform because insights stop at the dashboard instead of feeding back into formulation decisions. In that view, manufacturers create value when production data becomes formulation knowledge, allowing them to replace worst-case safety margins with evidence from actual plant performance. PetfoodIndustry’s separate coverage of digital quality tools made a parallel point: many production failures begin as data, versioning, supplier-change, or workflow errors upstream, and downstream QC can detect them but not undo them. (petfoodindustry.com)

Industry reaction has been notably pragmatic rather than evangelical. Sook described AI as a “trusted copilot” for operators, not a replacement for them, and said companies should start by seeing what data they already collect, comparing target versus actual moisture, and piloting one repetitive use case on one line. Another recent PetfoodIndustry report on automation echoed that caution, arguing that “AI for AI’s sake” is unlikely to pay off unless formulation, QA, and engineering teams align around practical tolerances and measurable production goals. (petfoodindustry.com)

Why it matters: For veterinarians and nutrition-focused teams, the significance is indirect but real. Manufacturing consistency affects whether the finished food matches the nutritional intent behind the formula, especially in dry diets where extrusion, drying, and coating influence moisture, density, texture, and palatability. Better process control can also reduce unnecessary giveaway and rework, which matters in a market still dealing with ingredient volatility and margin pressure. While clinicians won’t be choosing diets based on a plant’s AI stack, they do benefit when manufacturers can produce more consistent diets with fewer process-driven deviations. (petfoodindustry.com)

There’s also a useful caution here for the veterinary channel. AI doesn’t erase the need for sound formulation, validated nutrient data, supplier oversight, or quality systems. In fact, the reporting suggests the opposite: weak data and fragmented workflows are exactly what limit AI’s value. For companies selling science-led nutrition to veterinary teams and pet parents, that means credibility will hinge less on whether they use AI and more on whether they can show tighter process control, stronger traceability, and more reliable batch performance. (petfoodindustry.com)

What to watch: Over the next few years, look for AI adoption in pet food to stay concentrated in extrusion, batching, QC, and formulation support, where plants can tie improvements to waste reduction, startup time, and consistency. The next milestone will be whether manufacturers can move from isolated prediction models to true closed-loop systems that automatically inform the next formulation and, eventually, support faster quality and compliance decisions. (petfoodindustry.com)

Common questions

  • How is AI being used in pet food manufacturing?
    The article says AI is most useful when it connects formulation, production, and quality data, rather than acting as a standalone prediction tool.
  • What did the Alltech case study show?
    A moisture prediction model built from four years of plant data improved in-tolerance performance from about 60% to 87%, cut startup stabilization time by 43%, and was projected to save about $567,000 annually across two extrusion lines.
  • What problems are pet food plants trying to solve with AI?
    The article says plants still rely too much on operator judgment, have delayed QC feedback, and use siloed systems that do not create a true feedback loop between formulation, production settings, and QC results.
  • What should pet food companies do first with AI?
    They should start by reviewing the data they already collect, comparing target versus actual moisture, and piloting one repetitive use case on one line.

Like what you're reading?

The Feed delivers veterinary news every weekday.