Zero typing errors on batch and date

The problem

One digit typed on the coder can hold half a run.

When the work order changes, someone types the new batch by hand on the coder or body-welder panel. One typo can immobilize half a run at the packer client's plant.

• The coder that prints batch and date on each can's base has a documented API, and it still sits isolated from the ERP.
• At every work order change, somebody types the new batch by hand on the coder or body welder panel, with the same data already in the ERP. The same value is entered twice, and the second time it is entered under time pressure.
• Plants live with 2-3 typing errors a month, and some of them end in a hold.
• A wrong batch is rarely caught at the plant: it turns up at the packer client's plant, where it can immobilize half a run.

How it fits the IRIS system

Connect stitches ERP and coder through their APIs — neither system is replaced.

Connect stitches two modern systems together via API. When the ERP marks a work order start, it automatically publishes batch and varnish spec to the coder or welder. Changing the work order = one tap in the ERP, not typing on the machine.

Changing a work order becomes one tap in the ERP instead of typing on the machine. The batch that reaches the can is the batch the ERP holds, by construction. There is no second keyboard where the value can change on its way to the can end.

See the full IRIS architecture →

Before and after

The coder typed by hand versus the coder fed by the ERP

Aspect	Today	With iLEAN Connect
Batch and date at an order change	Typed on the machine panel	Published by the ERP at order start
Human-origin coding errors	2-3 a month	Zero
Varnish spec on the body welder	Selected by hand	Sent with the work order
Where a wrong batch is found	At the packer client	It cannot be produced
Order change on the line	Typing under time pressure	One tap in the ERP
Coder and ERP in metal packaging	—	Both kept, now talking to each other

2-3 typing errors a month (some tied to holds) → zero human-origin coding errors.

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 3-6 months, from cutting human-origin coding errors to zero.
• Holds caused by a mistyped batch or date stop happening, including those found only at the packer client. Each of those holds costs far more than the minutes saved at the order change.
• Order changes stop depending on somebody typing correctly under time pressure. The order change becomes faster too, because nobody has to look up and type the new values.
• And the varnish spec reaches the body welder with the work order, not from memory.

Estimated payback of 3 to 6 months, cutting human-origin coding errors to zero. Estimate to validate.

And the fair question from the production manager

“Isn't it simpler to just be careful when typing?” — the plant already is, and it still gets 2-3 errors a month: at cadence, care does not scale. Here the ERP value travels as is, with nothing reinterpreted on the way; the only interpretive step in the chain is upstream, when a work order is captured from paper, an anchored task where the best models drop below 1.5% error [1] and which a person validates before it reaches the ERP. What disappears is the keystroke on the machine, which is exactly where the 2-3 monthly errors come from.

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