Three machines stitched to the ERP — one tap changes the SKU on all three

The problem

The SKU change is where traceability is at stake, not steady-state production.

In a dairy processing plant making cheese in different formats and designations, the slicing and packing line in steady state is stable: the coder, the dynamic checkweigher and the slicer run alone, package after package. The problem is the SKU change, happening several times per shift when a new batch, a new reference weight or a new customer comes in. The three machines are modern — all have APIs — but nobody has stitched them to each other or to the ERP:

• The three machines concentrate the recall risk — a batch, designation, weight or allergen data point mistyped on any of the three panels can end in a recall of product already on the shelf.
• Typing spread across three panels at every change — someone enters by hand the batch and expiry into the coder, the designation and target weight into the checkweigher, and the cutting parameters into the slicer. Three screens, three chances to err, no cross-check between them.

The work order arrives from the ERP, but stays there: the ERP knows which SKU is due, but does not tell the three machines. Someone has to translate it by hand, three times, with the line stopped waiting.

How it fits the IRIS system

Connect stitches the three machines to the ERP via API — one tap, three panels in sync.

The SKU change needs no more screens or typing: the data that already lives in the ERP needs to reach the three machines at the same time, with nobody rewriting it three times. That is what Connect is for.

When the work order starts, the ERP publishes batch, designation, expiry date and allergens to the coder and the checkweigher; it publishes the cutting parameters to the slicer. One tap in the ERP, 30 seconds and the three machines are synchronized.

How Connect operates on the SKU change in a dairy processing plant:

• Connect stitches the three systems via API — the coder, the dynamic checkweigher and the slicer already have modern interfaces; Connect connects to all three and to the ERP without touching the line's PLC or replacing any equipment.
• The ERP publishes when the work order starts — batch, sales designation, expiry date and declared allergens reach the coder and the checkweigher the very instant the work order is started.
• The cutting parameters reach the slicer — thickness, number of slices and target weight per tray are published to the slicer without the operator re-entering them.
• A cross-confirmation before releasing the line — Connect checks that the three machines have received and applied the same SKU before the operator releases the first package; if one lags behind, it holds and alerts.
• A record of the change for traceability — which SKU was published, at what time, to which three machines and with which confirmation is archived, with nobody having to reconstruct it for the traceability file.

See the full IRIS architecture →

Before and after

A SKU change typed by hand vs. a SKU change synchronized with Connect

Aspect	SKU change typed by hand	SKU change synchronized with Connect
SKU change time	15-20 min of typing spread across three panels	30 s synchronized from the ERP
Typing errors	2-3 typing errors/month	Zero of human origin
Batch and expiry coding	Typed by hand into the coder	Published by the ERP when the work order starts
Designation and weight on the checkweigher	Typed by hand, no cross-check with the ERP	Published by the ERP along with the batch
Cutting parameters on the slicer	Adjusted by hand by the operator	Published automatically from the ERP
Confirmation across the three machines	None — each panel is an island	Connect confirms all three before releasing the line

Impact estimate

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

• Dairy processing plant with a coder, a dynamic checkweigher and a slicer already equipped with APIs, and several SKU changes per shift.
• Connect pilot on the SKU change — it stitches the three machines to the ERP via API, without replacing equipment or touching the line PLC. First value expected within a few weeks.
• Estimated payback between 3 and 6 months, depending on the frequency of SKU changes per shift and the average cost of a recall or rework. Estimate to be validated.
• Beyond the payback, SMED between batches becomes more effective: fewer stopped minutes per change, more clean startup windows.
• The hard lever is a single recall avoided from a data point mistyped on any of the three panels — one recall pays for the pilot several times over.

And the fair question from the production manager

"What if Connect goes down and we still have to start?" — Connect is not a single point of failure for starting production: if the integration is unavailable, the operator goes back to the usual manual typing, exactly as today. Connect adds no new dependency to produce; it removes redundant work when available and steps aside without blocking the line when it is not.

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