Zero labeling recalls — the iLEAN agentic system.
The number-one cause of recalls in the food industry is not a manufacturing failure: it is an information failure on a well-made product. A label from the previous SKU, a date from the previous shift, an allergen declaration that does not match the formulation. iLEAN closes that gap with a coordinated agentic system: the line does not release product until every package in the batch has been verified by the combination of Connect, Edge and Agents. The person signs.
The most silent and most expensive failure in any food plant.
It is the most silent and most expensive failure in any food plant, and it can happen to all of them: the product has been manufactured perfectly, but it reaches the shelf with the wrong label, a misprinted expiry date, an incorrect batch or an allergen declaration that does not match that formulation. It is not a manufacturing failure: it is an information failure, and it is today the number-one cause of product recalls in the sector.
It happens, above all, during SKU changeovers, when the outgoing format crosses paths with the next one: a label reel that was not changed in time, a date code left over from the previous shift, a correct product inside a package that announces something else.
The classic system (an attentive operator + a paper checklist + the shift lead's eyeball review) works 99% of the time. And that 1% is what ends up as a RASFF alert, a shelf recall, a call from the retailer at 10:00pm. The cost is not the fine: the cost is the recall, the rework, the reverse transport, the destruction of well-made product and the brand damage that appears on no invoice but weighs for years.
The iLEAN agentic system, step by step, inside a food plant.
The practical case shows how a coordinated AI system guarantees that every package leaving the line carries printed and declared exactly what it should. Step by step, the full sequence of the system:
- Batch startup — Connect capture. When the batch starts, the operator takes a photo of the production order document about to run on the line and a photo of the first product label to be printed for the whole batch. The AI digitizes both documents and stores them in central memory.
- Consistency check — central AI + personal assistant in the earpiece. The central AI verifies that the production order and the printed label match, and guides the operator step by step, through a personal AI assistant in the earpiece, to review and confirm one by one the critical points of that specific label.
- 100% in-line inspection — Edge + JIDOKA AI. The AI terminals installed on the line inspect every package by vision (dozens of times per second) and verify that the label, the expiry date, the batch and the allergen information are correct, automatically extracting any non-compliant unit (JIDOKA AI) without stopping the line, or even halting it if the error is repetitive or a manager requests it through their personal assistant in the earpiece.
- Traceability into the ERP — legacy integration. The AI integration system pushes each batch's traceability directly into the ERP (what was printed, on which unit and at what time), leaving a complete, auditable record, with no paper and no manual transcription.
- Assisted SKU changeover — SMED AI. The SMED AI system oversees the SKU changeover and verifies that it was executed to the defined standard, on time and in full, detecting any deviation the instant it happens, not days later.
- Legacy machine capture — graduated Connect. All of this process information is captured by AI even from old machines with no interface, and is automatically integrated with the company's legacy systems (ERP, MES, labelers, coders).
- Batch release — only if everything matches. The line does not release product for shipping until the system confirms that every package in the batch is compliant. The gap through which the vast majority of recalls slip simply ceases to exist.
Real result: this system eliminates rework, relabeling, stoppages at format changeovers and customer rejections at receiving — recovering hours of line time every shift — and, above all, closes at the root the exact point where most incidents originate: not a failure in manufacturing, but an information failure on a well-made product. It is not an insurance policy you might use once a year: it is measurable savings every day and, at the same time, the best possible prevention of the incident no plant manager wants to live through.
Classic labeling vs. the iLEAN agentic system
| Aspect | Line with classic control | Line with the iLEAN agentic system |
|---|---|---|
| Batch startup | Operator reads the work order, eyeballs the first label | Photo of the order + first label; AI cross-checks and guides the operator by earpiece |
| SKU changeover | Paper checklist, shift lead's signature | SMED AI oversees the sequence and catches the deviation instantly |
| Per-package verification | Manual sampling every N units | 100% vision on the batch, dozens of inspections per second |
| Non-compliant package | Reaches the end of the line, caught late or not at all | JIDOKA AI: automatic ejection without stopping the line |
| Traceability into the ERP | Manual transcription at end of shift, on paper or in Excel | Direct push — what was printed, on which unit and at what time |
| Old coders and labelers | "They can't be integrated, it costs more than it's worth" | Graduated Connect captures them (panel photo, serial reading, direct integration) |
| Batch release | On trust in the process | Only when the system confirms every package in the batch is compliant |
| Frequency of label-driven recalls | 1 in every N batches — the 1% that slips through to RASFF | Zero, by design of the system |
| File for IFS/BRC | Rebuilt by hand, weeks | Automatic per-batch dossier, with an image of every package |
Impact estimate for your plant — to be validated with your numbers.
The block below is an estimate to be validated with the specific data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.
- Food plant with 2-3 multi-SKU packing lines (salted, whole-grain, with sesame, allergen-free), with coders and labelers spanning several generations.
- Pilot on one line with the full combination — Connect capture at batch startup, Edge with JIDOKA AI on every package, SMED AI on the SKU changeover, ERP integration. First value in a few weeks.
- Indicative payback between 3 and 9 months. The hard lever is threefold: (1) daily savings on rework, relabeling and hours lost in format changeovers, (2) elimination of customer rejections at receiving, (3) a single avoided recall pays for the entire pilot — product pulled from the shelf, reverse transport, destruction and brand damage.
- Pilot line operational in 2-3 months; replication to the remaining lines executed by your own people, trained during the pilot, in the months that follow.
And the reasonable doubt: "what if the system itself gets it wrong?"
The quality manager's right question. Two answers that hold each other up. The technical one: hallucination is a problem of free generation, not of anchored tasks — reading a printed label and comparing it against the production order is as anchored a task as they come. In tasks of this type, the best models brought error below 1.5% [1]. The architectural one: the iLEAN system is built on the three IRIS safety rings — Connect transports, the Agents decide, the person signs. The critical verification never runs alone. It is not that everything goes through a person; it is that the system allows it where it matters, and here it matters.
[1] OpenAI paper "Why Language Models Hallucinate", 2025 — reliability of AI in anchored tasks.
What people ask about the iLEAN agentic labeling system
What percentage of food industry recalls are caused by incorrect labeling?
According to RASFF (the European food alert system) and AESAN, undeclared or misdeclared allergens are, year after year, the first or second leading cause of food product recalls in Europe — alongside microbiological contamination. When you add errors in dates, batches and misprinted codes, the label/information block far outweighs manufacturing failures. The paradox: the product is well made, the package lies. That is why we say it is not a manufacturing failure, it is an information failure — and why the savings from closing this gap are among the largest a food plant can capture.
Can I install only part of the system, or does it have to be everything?
You can start with the piece that hurts most and grow from there. The most common starting points are Edge verification on the line (a camera that reads label + date + batch and applies JIDOKA AI to every package) and Connect capture at batch startup (a photo of the production order and of the batch's first label). With those two pieces you already close 80% of the risk. The SMED AI for the SKU changeover and the ERP integration come in the next phase. iLEAN is sold modular precisely because the Pareto rules — the ripe fruit comes first.
Does it work with coders and labelers of any brand?
Yes. iLEAN does not force you to change your coder or your labeler — it works on top of the ones you already have. Edge sees the package and the label the way a person would: with an industrial camera over the line itself. And Connect captures the coder's data with three degrees of integration depending on what the machine allows — from a photo of the panel of an old coder with no interface, through serial reading on intermediate equipment, to direct integration with modern coders. That is the system's commitment: we do not force you to throw away anything you already have.
How long does it take to be operational in a plant?
The standard iLEAN method — a 3-5 day immersion with a mixed team (your plant's people + embedded industrial-AI engineers), Pareto applied, first value in a few weeks, a full pilot line typically in 2-3 months. The rollout to the remaining lines is replicated by your own people, trained during the pilot. The most documented real case (powder coating, automotive) had first value in 2 weeks, a complete pilot in 60 days, and the client expanded to the other 4 lines with its own people in 3 months, without the vendor. That is what we call a pilot that creates no dependency.
How do you guarantee that the system itself does not get it wrong?
Two answers, one technical and one architectural. Technical: hallucination is a problem of free generation, not of anchored tasks. Reading a label and comparing it against the production order is an anchored task — the best recent study (OpenAI paper, 2025) puts the error of the best models on anchored tasks below 1.5%. Architectural: the iLEAN system is built on the three IRIS safety rings — the critical verification never runs alone, there is always a human who signs. Connect transports, the Agents decide, the person signs. That is the golden rule and it does not break.
Tell us your case and in 48h we'll send you the estimated ROI of this agentic system for your food plant.
We work on your plant's real data, not ours. Diagnostic with no commitment.
Request estimated ROI in 48h ‹ All food product cases See food industry