Zero typing error on the final label

The problem

The last step before the product leaves the plant is typed by hand.

The coder prints batch, expiry and registration number on every bottle or drum. It's the last physical step before the product leaves the plant. The coder is 5 to 15 years old and already has an API, but it's isolated from the ERP: on every batch change, someone types the new data by hand into its panel. A mistyped digit means a container with the wrong batch or expiry printed: not legally sellable, and a traceability risk if it reaches the field.

• The coder prints batch, expiry and registration number on every bottle or jerrycan. It is the last physical step before the product leaves the plant.
• The coder is 5 to 15 years old and already has an API, but nobody connected it to the ERP. At every batch change, someone types the new data on its small panel, often against the clock and with the filler already running.
• One mistyped digit means a container with the wrong batch or expiry: not legally sellable, and a traceability risk if it reaches a grower's field.
• And the error is usually found by a customer or an inspector, not on the line, because the print looks perfectly correct. By then the whole batch is in distribution and the only fix is to recall and recode.

How it fits the IRIS system

Connect in integration mode — the coder and the ERP stitched through their APIs.

iLEAN Connect stitches the coder to the ERP via API: when the ERP marks a work order as started, it publishes batch, expiry and registration automatically. Changing batches is one tap in the ERP, not typing on the coder.

Nothing is replaced: the coder already speaks, it just was never asked. When the ERP marks a work order as started, the batch, expiry and registration are published to the coder automatically. Changing batch becomes one tap in the ERP instead of typing on a small panel at the line, with gloves on and the filler waiting.

See the full IRIS architecture →

Before and after

The isolated coder versus the coder fed by the work order

Aspect	Today	With iLEAN Connect
Source of batch and expiry	Typed on the coder's panel	The active work order
Registration number	Typed, or left from the last product	Published with the batch
Batch change	Typing at the line	One tap in the ERP
Mistyped digit	A container that cannot be sold	Zero typing error
Where the error is found	At the customer or the inspection	It is not produced
The coder itself	—	Kept, through its existing API

Before: on every batch change someone types batch, expiry and registration into the coder panel, and one wrong digit leaves unsellable containers. After: a batch change is one tap in the ERP and the final label prints with no typing error.

Impact estimate

Impact estimate — to be validated with your format change frequency.

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.
• Zero typing error on batch and expiry on the final label, because nobody types them any more: they come from the same work order that drove the formulation.
• Batch changes at the bottling line become faster and stop depending on who is at the coder's panel, or on whether they remembered to change all three fields.
• And the registration number printed always matches the product actually being filled, which is the detail an inspector checks first and the one a grower relies on to know what they are applying.

Estimated payback 3-6 months · zero typing error on batch and expiry on the final label.

And the fair question from the production manager

“What if the ERP sends the wrong data to the coder?” — the coder prints what the work order says, so the source of truth is a single record already confirmed by a person, not a second copy typed at the line. The only extraction involved is checking the printed code back against the order, an anchored task where the best models drop below 1.5% error [1]; a mismatch stops the change before the first container is coded, not after a pallet has been filled.

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