In sequence you can stop the customer's line in two ways

The problem

In sequence there are two ways to stop the customer's line, and they cost the same.

An incident that reaches the customer's line costs, between direct charge, containment and commercial consequences, from tens to hundreds of thousands of euros per event. And it drags along controlled shipping, sorting paid to third parties, an extraordinary customer audit, a drop in supplier ranking and, at worst, a program that is not renewed. In sequenced delivery there are two distinct ways to cause it: sending a bad part, or sending a good part in the wrong order. They cost the same on the customer's line but they are different technical problems, and a defense covering only one of them leaves half the risk open. It is the risk the operations director carries every day, and the one that no isolated investment in a machine or a control fully solves.

• An incident reaching the customer's line costs, between direct charge, containment and commercial consequences, from tens to hundreds of thousands of euros per event.
• And it drags along controlled shipping, sorting paid to third parties, an extraordinary customer audit, a drop in supplier ranking and, at worst, a program that is not renewed.
• There are two ways to cause it: sending a bad part, or sending a good part in the wrong order. They cost the same but they are different technical problems.
• A defense covering only one of them leaves half the risk open.

How it fits the IRIS system

The four coordinated rings — and the last one is at expedition.

Four coordinated rings.

If the four rings do not agree, the rack does not leave the plant. Facing a claim, the answer takes hours and comes with data, not days and consensus.

• Ring 1 · Process: the digitized press curve and the variant changeover validation guarantee the part is born with the right tooling and parameters. JIDOKA AI does not allow startup otherwise.
• Ring 2 · Part: the Edge cameras verify surface, trim and presence of every component of that specific variant before the rack. The bad part leaves the sequence, not the plant.
• Ring 3 · Identity and order: the labeler prints what the sequencing system dictates, the scale checks the rack weight and customer change alerts arrive within seconds. The good part goes in the right position.
• Ring 4 · Evidence: every rack ships with its dossier, cross-referencing the three previous rings with the signed human validations. Facing a claim, the answer takes hours and comes with data.
• SMED AI speeds up the changeover between variants so the four rings are not paid for in productivity, which is the shop floor team's legitimate objection.

See the full IRIS architecture →

Before and after

Incident found outside vs. discrepancy stopped inside

Aspect	Without the four rings	With the coordinated system
Where it is detected	On the customer's line	Inside the plant, before expedition
Cost of the event	Tens to hundreds of thousands per incident	Moving one rack
Bad part	The customer's operator stops it	It leaves the sequence on the line
Good part in the wrong order	Nobody catches it until the customer	Label and scale cross-check it
Facing a claim	The charge is argued with no data	Bounded with traceability in hours
Supplier ranking	Exposed on every shipment	Protected systematically

From incidents detected by the customer to discrepancies detected inside the plant, before expedition. From arguing a charge with no data to bounding it with full traceability in hours.

Impact estimate

Impact estimate for your plant — to validate against your history.

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.

• Plant delivering in sequence to one or several customer plants, with a hundred active variants.
• Indicative payback between 6 and 12 months against the first avoided incident.
• The value of protecting supplier ranking and program renewal is multi-year and far greater than that of the isolated incident.
• To validate against the last three years' history of charged incidents, which is the number that really sizes this case.

Estimated payback 6 to 12 months against the first avoided incident. The value of protecting supplier ranking and program renewal is multi-year and far greater than that of the isolated incident. *Estimate to validate against the last three years' history*.

And the fair question from the production manager

“What if the system holds a rack that was fine and stalls expedition?” — the hold happens when the rings do not agree, and the lead sees the specific discrepancy on their tablet and can resolve it on the spot. Compare it with the alternative scenario, which is not smooth expedition: it is that same rack reaching the customer's line. A hold costs minutes; a sequence incident costs a charge.

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