The line does not start until the brand changeover is clean

The problem

The changeover is well designed. The problem is how it gets signed.

Line clearance between brands is manual, done under shipping-calendar pressure, and its only evidence is a signed checkbox. The risk is not theoretical: in a plant bottling for several brands, packaging materials from different clients live in the same warehouse. If a single label from the previous brand gets through, the problem is not a defect: it is a misidentified product in the market, with a brand client and a certifying body involved. An incident like that is not settled with a credit note. It puts the whole contract-production agreement in question — and that is multi-year revenue.

• Between one brand and the next, the line has to be cleared of every bottle, label, back label, capsule and marbete from the previous one. Today that rests on a person walking the line and signing "line cleared", under shipping-calendar pressure, with no evidence beyond their word.
• In a plant bottling for several brands, packaging materials from different clients live in the same warehouse. The risk is not theoretical: a pallet of the wrong back labels next to the right ones is an ordinary Tuesday.
• If a single label from the previous brand gets through, it is not a defect: it is a misidentified product in the market, with a brand client and a certifying body involved.
• An incident like that is not settled with a credit note. It puts the whole contract-production agreement in question, and that is multi-year revenue.

How it fits the IRIS system

Edge with JIDOKA AI — the line does not start without conformity and a signature.

Edge + JIDOKA AI. The flow:

The cameras do not decide alone: each point is compared against the reference state for the incoming brand — which also covers what should be there, not just what should not — and it is the quality manager who signs on that evidence. What changes is not who decides, but what they decide on. The line does not run slower for being safer: the check is seconds, and it replaces a walk nobody fully trusted.

• Industrial cameras at each clearance point: bottle hopper or depalletiser, labeling head, capsule applicator, tax stamp applicator and case packing table.
• The network compares the current state against the reference for "cleared and prepared for the incoming brand" — which also covers which materials should be there, not just which should not.
• If any point fails, the line does not start and the system points out exactly which station and what it found.
• Start-up requires all N cameras green plus the quality manager's signature.
• Before-and-after photos of every point are stored and folded into the batch evidence pack (case 11).

The machine does not decide alone, and the person no longer signs blind. The three safety rings, active at the plant's highest-risk moment.

See the full IRIS architecture →

Before and after

Today's changeover versus the validated changeover

Aspect	Today	With iLEAN Edge
Line clearance	A signed checkbox	N cameras green plus a signature
A label from the previous brand	Found in the market	Blocked before start-up
Which station failed	Nobody knows	Pointed out, with what it found
Evidence of the clearance	Somebody's word	Before-and-after photos per point
The quality manager	Signs blind	Signs on evidence
Start-up time	Under calendar pressure	Seconds of check, no extra margin

From an unevidenced signature to one backed by visual evidence point by point. From the risk of cross-brand material to a block before start-up, with before/after photos stored.

Impact estimate

Impact estimate — to be validated with the quality manager.

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.

• Protection of the contract-production agreement and of the relationship with each brand client, whose multi-year value dwarfs any single incident.
• Estimated payback 6-12 months, depending on brand changeovers per month.
• Cross-brand material caught before start-up instead of in the market.
• And a clearance record with before-and-after photos per point, folded into the batch evidence pack.

Protection of the contract-production agreement and of the relationship with each brand client, whose multi-year value dwarfs any single incident. Estimated payback 6-12 months depending on brand changeovers per month. *Estimate to be validated with the quality manager.*

And the fair question from the production manager

"What if the cameras block the line for no reason?" — that is the right objection, because the changeover sits on the shipping calendar. Conformity is checked against the reference state for that specific brand, an anchored task where the best models drop below 1.5% error [1], and in case of doubt the system does not block silently: it names the station and shows what it saw, and the quality manager decides. It stops on a verifiable discrepancy, not on model uncertainty. The quality manager can override with a signature, and the override is stored with the photo of what the camera saw.

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