THE CODER STOPS BEING HAND-TYPED

The problem

A modern machine, a documented API, and somebody typing into it anyway.

The coder that prints batch and date on every sachet or cube wrapper is relatively modern — it has an API — but lives isolated from the ERP. When the format or brand on the same line changes, someone hand-types the new batch and allergens on its panel: one mistyped digit contaminates half the run with wrong data.

• The coder that prints batch and date on every sachet and cube wrapper is recent equipment with a documented interface, and it sits completely outside the ERP's flow. It was bought connected and installed isolated.
• Every format or brand change on the same line means somebody hand-types the new batch and allergen line on its panel, copying data that already exists three meters away in the production order.
• One mistyped digit contaminates half a run with wrong data, and the run is cartoned, palletized and shrink-wrapped long before anybody notices.
• On a line where brands alternate all day, that exposure repeats at every changeover, all year round, and it only takes one of them to go wrong.

How it fits the IRIS system

Connect stitches coder and ERP by API — neither system is replaced.

Connect stitches the coder to the ERP via API: when the ERP marks the start of a production order, it automatically publishes batch, date and the active recipe's allergen list to the coder, with no manual typing in between.

What was actually printed comes back to the system as confirmation, not as an assumption. That difference is what turns the printed code into evidence you can put in front of an auditor or a distributor, instead of a well-founded belief about what should have happened on the line.

See the full IRIS architecture →

Before and after

The isolated coder versus the coder fed by the order

Aspect	Today	With iLEAN Connect
Typing batch and date	Once per changeover, by hand	Never
The printed allergen line	Whatever was left on the panel	Published from the active recipe
A single mistyped digit	Half a run coded wrong	Not possible by construction
Changeover time at the panel	Minutes on the critical path	The order start in the ERP
Proof of what the sachet carries	Assumed from the panel	Returned by the machine itself
The coder and the ERP	—	Both kept exactly as they are

Monthly typing errors, some tied to recalls → zero human-origin coding errors.

Impact estimate

Impact estimate — to be validated with the packing line's own data.

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 3-6 months, counting only the panel time recovered across every changeover in the shift.
• Human-origin coding errors go to zero, and with them the most common single cause of a labeling or undeclared-allergen recall.
• The allergen line stops depending on whoever last touched the panel and starts depending on the recipe quality approved, which is where it belonged all along.
• And a run coded with the previous brand's data, allergen line included, stops being something that can happen at all on that line.

Estimated payback of 3 to 6 months, with a drastic reduction in labeling/allergen recall risk. Estimate to be validated.

And the fair question from the production manager

"Isn't a proper integration project cleaner than this?" — it is, the day it reaches the top of the IT list, and that is exactly the problem: this machine has had an API for years and is still fed by hand because the window never came. The stitching replaces neither system and needs no shutdown. And where a value has to be read back from the machine rather than published to it, the task is anchored and the best models stay below 1.5% error [1], with the line tablet behind it as the second net. The case is measured in weeks, not in the quarters an integration project takes.

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