You change UV inkjet ink every shift — and the changeover eats the next customer's hour.

The problem

The ink changeover is where a multi-customer plant wins or loses the shift.

In multi-customer packaging the plan fills up with short jobs: one run for a cosmetics customer with a spot color, another for a food contact customer with a certified primary, another for a pharma customer with a serialized code. Between each job, an ink changeover. And that changeover touches four islands:

1. The next job's recipe — in the ERP, with its colors, substrate adhesion and certification requirements.
2. The circuit purge — a procedure the operator follows from memory or from a laminated sheet next to the machine.
3. The printhead — temperature, pressure, pulse configuration in the machine HMI.
4. QA — a print trial sent to the lab, then the wait for a reply, and only then does the production run start.

The operator jumps between three screens, the ERP and the laminated sheet. The theoretical 18-minute changeover becomes 35 when the lab is swamped or the recipe arrives with a new note from the customer. And by the end of the shift that is two extra changeovers eating into the next customer's hour.

How it fits the IRIS system

iLEAN does not change your inkjet machine — it seals the gap between the machine, the ERP and the lab.

The fast-changeover problem is not a lack of procedure: it is procedure scattered across four islands that the operator has to reassemble in their head at every changeover. iLEAN acts as the putty that seals those gaps without asking you to change your machine, your ERP or your LIMS.

Connect carries. Agents decide. The person signs. The changeover becomes a single guided flow.

The iLEAN pieces applied to SMED for inkjet ink in packaging:

• Connect — captures the recipe from the ERP at second zero, reads the machine HMI whether it comes from an old or a modern panel, picks up the purge procedure from whatever document already existed and presents it to the operator on their tablet or earpiece. It talks both ways: the lab confirms QA through the same channel. Whatever used to arrive by email from the customer with a special note comes in too, with no forwarding.
• Orchestration agent — sequences the four steps according to the previous and next color: if you go from black to white, purge A applies; if you go from white to magenta, it is purge B. It proposes printhead temperature and pressure based on the recipe. It sends the sample to the lab with the job's QR code, not in an anonymous tray.
• QA agent — cross-references the trial's color curve against the customer's, proposes approval/rework to the QA lead, and the lead signs from their own channel. The operator sees the green light on their tablet and starts. Every signature stays traceable in the batch dossier.

See the full IRIS architecture →

Before and after

Manual changeover across three screens vs. a guided changeover with iLEAN Connect

Aspect	Manual changeover + checklist	With iLEAN Connect + Agents
Source of the recipe	ERP, copied into the operator's notebook	Recipe reaches the flow at second zero
Purge procedure	Laminated sheet next to the machine	Procedure matched to the previous-color/next-color pair
Printhead setup	Operator types it into the HMI	Temperature and pressure proposed; operator confirms
Sample to the lab	Anonymous tray, phone call	Job QR code, QA approves/reworks through the channel
Changeover traceability	Rebuilt from shift log + incident report	Batch dossier with every step and every signature
Changeovers when the lab is swamped	Queue on the floor waiting for the lab	Prioritization proposed, the person decides

Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Packaging plant with 2-6 UV inkjet machines, 10-20 changeovers a day, a multi-customer mix with food contact and occasional pharma.
• Connect pilot on the Pareto machine (the highest number of changeovers or the most time lost in QA). First value expected within a few weeks.
• Expected changeover time reduction ≥ 30%, with a defensible floor in minutes per changeover recovered.
• Indicative payback between 4 and 9 months, depending on the number of daily changeovers, the machine hourly cost and the penalty for a job delivered late.
• The hard lever is the hour gained per shift (which gets sold to another customer) and the QA dossier ready for audit with no extra manual work.

And the operations director's reasonable doubt

“What if the agent proposes the wrong purge and wrecks the printhead?” — hallucination is a problem of free generation, not of anchored tasks. The purge proposal is anchored to the previous-color/next-color matrix your own team has already written down; the agent does not invent, it recontextualizes. In anchored tasks, the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: the operator confirms before executing and QA signs before the production run starts.

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