The line doesn't start without a camera green light

The problem

The changeover is signed on someone's word.

today the changeover sign-off rests on a human's word with no structured visual evidence, and one leftover bundle can contaminate an entire customer order.

• Every size-color changeover requires checking that no piece or bundle of the previous reference is still on a station, a cart or under a machine. On a sewing line with dozens of stations, that is a lot of places to look in a hurry.
• Today that check rests on a human's word, with no structured visual evidence behind the sign-off.
• One leftover bundle of the previous color or size is enough to contaminate an entire customer order. And it is usually discovered by the customer, when the carton is opened.
• And on made-to-order corporate uniforms, with many short runs, changeovers multiply and so does the exposure. Each changeover is a new chance for one bundle to stay where it should not.

How it fits the IRIS system

Edge with JIDOKA AI — the line waits for every camera's green light.

Edge + JIDOKA AI. Industrial cameras with CNNs trained to recognize "line clear vs. residual bundle". The line doesn't start until every camera confirms and quality signs off backed by visual evidence.

The cameras compare each station with its reference state and quality signs off on that evidence. What changes is not who decides, but what the sign-off rests on. A sign-off backed by images is quick to give and hard to dispute later, by the plant or by the customer.

See the full IRIS architecture →

Before and after

Today's changeover versus the camera-cleared changeover

Aspect	Today	With iLEAN Edge
Changeover sign-off	A person's word	Backed by images at every critical point
Leftover bundle from the previous reference	Found at packing, or by the customer	Detected before startup
Startup	When someone says so	When every camera confirms
Evidence after the fact	None	Before/after images per changeover
Short made-to-order runs	More changeovers, more risk	More changeovers, same check
Doubtful case	—	Escalated to quality

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

Impact estimate

Impact estimate — to be validated with your changeover calendar.

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.

• Payback varies with the number of changeovers per shift, so we do not fix a range here: we estimate it with your changeover calendar.
• The value is protecting high-value customer programs that carry contractual penalties for spec non-compliance. One avoided penalty can be worth more than the whole installation.
• The sign-off moves from blind to backed by visual evidence at every critical point. Quality signs on what it sees, not on what it is told.
• And the leftover bundle is caught at the station, not in a mixed carton. And the evidence of every changeover is captured once and reused in audits.

protects high-value customer programs with contractual penalties for spec non-compliance, payback varies with changeovers per shift. *Estimate to validate*.

And the fair question from the production manager

“What if a camera blocks startup for no reason?” — that is the right objection, because a changeover sits on the line's critical path. Recognizing a clear station against one with a residual bundle, at a known spot, is an anchored task, where the best models drop below 1.5% error [1]. And when the doubt is the model's, JIDOKA AI does not block silently: it escalates to quality, who decides on the image. The line waits for a person, not for a model.

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