Export configuration, verified before packing

The problem

One line, dozens of electrical codes, and a checkbox as the only evidence.

Exporting to more than 50 countries means the same line builds units with different voltage, plug and regulatory label depending on destination. If the configuration changeover between one lot and the next isn't verified well, the unit reaches the wrong country with the wrong voltage. Today it relies on a human confirming "correct configuration" with no objective evidence, on a line that can change destination several times a shift.

• Exporting to more than 50 countries means the same line builds coolers with different voltage, plug and regulatory label depending on destination. A cooler for a 120 V, 60 Hz market and one for a 230 V, 50 Hz market can look identical from a few meters away.
• The line can change destination several times a shift. Each changeover means swapping the plug, the rating plate and the label set.
• If the changeover between one lot and the next is not verified, the cooler reaches the wrong country with the wrong voltage.
• Today that verification rests on a person confirming “correct configuration”, with no objective evidence behind the tick and the line waiting to restart. The error is not discovered at the plant; it is discovered by the importer, or worse, at the point of installation.

How it fits the IRIS system

Edge with JIDOKA AI — no cooler goes to packing until the camera and the line lead agree.

Edge plus JIDOKA AI: cameras at the configuration checkpoint compare the current state — plate, plug, label — against the reference state for the destination country on the active production order. The unit isn't released to packing until the camera clears it and the line lead signs off.

The camera does not replace the line lead's sign-off; it gives it something to rest on. The plate, the plug and the label are compared with what the order's destination requires, and the line lead signs on visual evidence instead of on memory. It is jidoka in its original sense: the line stops on the abnormality, and a person decides with the evidence in hand.

See the full IRIS architecture →

Before and after

The export changeover today versus the verified export changeover

Aspect	Today	With iLEAN Edge
Voltage rating plate	Assumed correct	Read and compared with the order
Plug type	Checked by eye, sometimes	Checked on every unit
Regulatory label	Taken from the right box, hopefully	Matched to the destination
Line lead sign-off	A tick with no evidence	Signed on visual evidence
Several destinations per shift	More changes, more exposure	Same check on every change
Wrong-country shipment	Discovered by the importer	Held before packing

Blind sign-off with wrong-country shipment risk → sign-off backed by visual evidence per unit.

Impact estimate

Impact estimate — to be validated with your numbers.

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.

• Estimated payback 6-12 months, depending on the weight of the multi-country export mix.
• Target: zero shipments to the wrong country because of voltage configuration. A cooler sent to the wrong market means freight both ways, rework and a customer who loses trust in the plant.
• Blind sign-off with wrong-country shipment risk becomes sign-off backed by visual evidence per unit.
• And every unit carries the image of its plate, plug and label, which is useful later for the importer and for certification. That image is also proof for the destination's certification if it is ever questioned.

Estimated payback of 6-12 months depending on the weight of the multi-country export mix. Estimate to validate.

And the fair question from the production manager

“And when the line is behind, the camera will become the bottleneck.” — the check takes seconds per unit and only stops the cooler on a verifiable mismatch between the plate or plug and the order's destination. Reading a rating plate and matching it to a closed list of voltages is an anchored task, where the best models drop below 1.5% error [1]; when the doubt is the model's, it escalates to the line lead instead of blocking silently.

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