The Color Change Won't Start Without Proof

The problem

The paint shop's most frequent event is signed off with no evidence.

A color change is one of the most frequent events of the day in paint, and its validation depends on a human sign-off with no structured evidence. A small margin of error in the purge translates into repaint rework or, worse, a color-defective unit that goes unnoticed until delivery.

• In a plant that paints in customer-order sequence, the color changes many times per shift — sometimes every few bodies. Each change requires a certified purge of the nozzles and lines.
• If pigment from the previous color remains, the first bodies of the new color come out contaminated: a shade off, or specks of the old color in the new one.
• The purge is validated by a human sign-off, with the sequence pressing and nothing structured to look at.
• A contaminated body goes back for a repaint at best; at worst, a shade defect passes unnoticed until delivery inspection.

How it fits the IRIS system

Edge with JIDOKA AI — the new color does not start until the evidence clears.

Edge + JIDOKA AI: fixed cameras at each critical purge point compare the current state against the 'clean for new color' reference; JIDOKA AI blocks the new sequence from starting until every camera reads OK and the quality manager signs off.

Each critical purge point has its own camera and its own reference image of the clean, ready state. JIDOKA AI does not let the next color start until every camera reads OK and the quality manager signs, now with the evidence on the tablet. The decision stays human; it simply stops being blind. And because the check is the same on the first change of the shift and the fortieth, fatigue stops being part of the equation.

See the full IRIS architecture →

Before and after

The color change signed blind versus signed on camera evidence

Aspect	Today	With iLEAN Edge + JIDOKA AI
Basis for the quality manager's sign-off	Judgment under sequence pressure	OK per critical purge point
Leftover pigment in a nozzle	Found on the first bodies	Detected before the next color starts
Start of the new color	Whenever someone signs	Blocked until every camera reads OK
Exposure across the shift	Every color change in the sequence	Every change, with the same check
Contaminated first body	Repaint, or a claim at delivery	Not painted
Record of each color change	A signature on a form	Photos, owner and time per point

Blind sign-off with color-contamination risk → sign-off based on visual evidence per critical point.

Impact estimate

Impact estimate — the payback scales with your color changes per shift.

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 depends on how many color changes you run per shift: we do not fix a range because it varies from plant to plant with color batching and sequence policy.
• Protection against color contamination on the first bodies after each change, which is where the repaint cost of this event sits.
• Sign-off based on visual evidence per critical point, instead of a blind signature given with the sequence pressing.
• And a record of every color change that later serves as evidence in the IATF 16949 process audit, without extra work.

Payback depends on color-change frequency per shift. *Figure to be validated*. (Payback varies by changeover frequency · protects against color contamination)

And the fair question from the production manager

“Won't the cameras hold the line for nothing at every color change?” — comparing a purge point against its own clean-state reference image is an anchored task, where the best models drop below 1.5% error [1]. JIDOKA AI blocks on a verifiable difference from that reference, not on model doubt; when the image is unclear, the point is escalated to the quality manager, who decides with the photo in front of them.

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