Allergen changeover, no gamble

The problem

The cleaning is validated by a signature that saw nothing in particular.

the list of declared allergens per product is long, and the responsibility for changeover cleaning falls on a human sign-off with no structured visual evidence. A poorly cleaned residue can contaminate the first batch of the next allergen.

• A line that alternates mayonnaise with egg, a mustard dressing, a dairy-based sauce and one with gluten needs validated cleaning before every allergen changeover.
• The list of declared allergens per product is long, and the responsibility falls on a human sign-off with no structured visual evidence behind it.
• Emulsions cling: a film of mayonnaise in the filler nozzle or under the doser is easy to miss when the line is behind schedule.
• A residue nobody photographed can contaminate the first batch of the next product, which then carries an allergen its label does not declare.

How it fits the IRIS system

Edge with JIDOKA AI — the line does not restart until every camera agrees.

Edge plus JIDOKA AI. Fixed industrial cameras at every critical point (hopper, doser, table, filler) compare the current state against the "clean to restart" reference. The line doesn't restart until every camera agrees and the quality manager signs off with that evidence.

The cameras do not sign. They compare each critical point with its clean-to-restart reference and show the quality manager what they saw. The person still decides; the decision now rests on evidence for every point, not on a box ticked at the end. Six months later, when a retailer asks how the first batch after a mustard run was released, the answer is a set of dated photographs, not a name.

See the full IRIS architecture →

Before and after

The changeover sign-off today versus the camera-backed sign-off

Aspect	Today	With iLEAN Edge
Basis for the sign-off	A signature on a checklist	Every critical point against its reference
Emulsion film in the filler nozzle	Found if someone looks	Holds the restart
Hopper, doser and in-line table	Checked in a hurry	Each under a fixed camera
First batch of the next allergen	Released on trust	Released on evidence
The sign-off six months later	A name and a time	Dated photographs per point
A week with many allergen changes	Exposure adds up	Each change closed with its proof

blind sign-off with cross-contamination risk → sign-off backed by visual evidence at every critical point.

Impact estimate

Impact estimate — to be validated with your allergen production mix.

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

• No payback range claimed here: the return depends on how much allergen switching weighs in your production mix, and we prefer not to invent months.
• What it protects is the relationship with customers and retailers that demand guaranteed allergen declarations, often the condition for keeping a private-label contract.
• Cross-contamination stops depending on how tired the crew is or how late the line is running.
• And every changeover leaves the evidence that case 11 assembles into the audit dossier, at no extra effort.

protects relationships with customers and retailers that demand guaranteed allergen declarations. *Estimate to validate based on how much allergen switching weighs in the production mix.*

And the fair question from the production manager

"Can a camera tell clean steel from a thin film of mayonnaise?" — the comparison is anchored: same point, same angle, same lighting, against a reference state approved by quality, where the best models stay under 1.5% error [1]. And the system never declares the line clean on its own: an ambiguous image holds the restart and goes to the quality manager. A false hold costs minutes; a false release costs a batch.

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.