Dosing Scale Tied to ERP

The problem

The scale has an API. The quantities are still typed by hand.

The main dosing scale already has an API but remains isolated from the ERP: every formula change requires manually typing quantities into its panel, risking an out-of-spec formula from a single mistyped value.

• The main dosing scale is modern and already exposes an API, but it was never connected to the ERP where the formulas and their approved versions live. It was bought for its accuracy, and its integration was left for later.
• Every formula change means an operator typing target quantities and tolerances into the scale panel, ingredient by ingredient, often from a printout of the formula.
• One mistyped digit is an out-of-spec batch: a fiber at ten times its dose, a protein short by a kilo, a salt tolerance entered in the wrong unit.
• Those errors show up as rejected batches, rework and customer complaints, and they happen every month no matter how careful the team is, because typing thirty values per changeover is exactly where people make mistakes.

How it fits the IRIS system

Connect stitches ERP and dosing scale together — the formula loads itself at batch start.

Connect stitches the ERP and the dosing scale together via API: when a formula's start is marked, the ERP automatically publishes target quantities and tolerances to the panel.

This is the case with the shortest path to value, because the hardware is already there: what is missing is the stitch. When the batch start is marked, the ERP publishes the target quantities and tolerances of that formula version directly to the panel, and the actual weights come back the same way. The operator's job moves from typing thirty numbers to confirming one formula, which is the kind of change people on the floor accept immediately.

See the full IRIS architecture →

Before and after

Formula change today versus formula pushed from the ERP

Aspect	Today	With iLEAN Connect
Loading a new formula on the dosing scale	Typed at the panel	Published from the ERP
Formula version the scale doses	Whatever was typed	The released version only
Manual dosing-entry errors per month	Recurring, with rejected batches	Zero of human origin
Actual dosed weights back to the ERP	Rewritten on paper	Returned automatically
Time to switch formula at the panel	Minutes of typing and checking	Seconds
Dosing scale hardware	—	Unchanged, existing API

Monthly typing errors (with rejected batches) → zero human-origin dosing 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 the rejected batches and rework avoided.
• Monthly manual typing errors at the dosing panel go to zero errors of human origin.
• Formula changes stop depending on how carefully someone types at the end of a long shift, and changeovers get shorter as a side effect.
• And the actual dosed weights flow back to the ERP, which closes the actual-versus-target loop that the weighing sheet case opens on the manual booths.

Payback 3-6 months · eliminates manual dosing-entry errors — estimate to validate

And the fair question from the production manager

"What if the ERP sends the wrong formula?" — then the problem is in the ERP, and that is exactly what this case exposes: the scale receives the released version only, and every push is logged with the formula version and who marked the batch start. Where iLEAN reads anything back — a dosed weight, a status — it is an anchored task where the best models drop below 1.5% error [1], and the operator still confirms the loaded formula on the panel before dosing.

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