The coder prints what the ERP says, with no typing

The problem

The same data typed twice, and the error only shows on the pallet.

The coder and the tax stamp control system are "almost integrated": they have interfaces, but nobody had time to tie them to the ERP, so they run as islands fed by keyboard. The bottling supervisor types at the start of every job and types again when the format or the market changes. Every keystroke is a chance for error, and the error only surfaces once the bottle is already cased: the right brand with the previous batch code, or the data set for the wrong market. On top of that comes tax stamp reconciliation, done today by hand and after the fact: adding up what was consumed and squaring it against bottles filled. By the time a discrepancy appears, weeks have gone by.

• The coder and the tax stamp control system are "almost integrated": they have interfaces, but nobody had time to tie them to the ERP, so they run as islands fed by keyboard.
• The bottling supervisor types brand, category, batch and destination market at the start of every job, and types again when the format or the market changes. The same data is already sitting in the ERP bottling order.
• The error only surfaces once the bottle is cased: the right brand with the previous batch code, or the data set for the wrong market. In a plant bottling several brands under the same standard, that is where the expensive mistake gets in.
• Marbete reconciliation is done by hand and after the fact, adding up what was consumed against bottles filled. By the time a discrepancy appears, weeks have gone by.

How it fits the IRIS system

Connect as a stitcher of modern systems — replacing neither.

Connect as a stitcher of modern systems. The flow:

What the ERP says is what the machine prints, with no intermediate typing, and the stamp range assigned to the order is registered at the start and reconciled at the close against the bottles the Edge camera actually counted. A coding error stops being something you catch on the pallet and becomes something that cannot happen by construction. And the bottling supervisor keeps doing the same job with one screen fewer.

• The ERP bottling order — brand, category (blanco, reposado, añejo, extra añejo, cristalino), batch, volume, proof, destination market — is published to the coder through its API.
• What the ERP says is what the machine prints. No intermediate typing.
• The range of tax stamps assigned to the order is registered at the start.
• When the order closes, it reconciles automatically against the bottles actually filled, counted by the Edge camera from case 9.
• Any discrepancy is visible the same day, not at month end.

Without replacing the coder, without replacing the ERP and without anyone learning a new screen.

See the full IRIS architecture →

Before and after

The isolated coder versus the stitched coder

Aspect	Today	With iLEAN Connect
Keying per bottling job	Twice	Zero
Brand, category, batch, market	Typed on the coder panel	Published from the ERP order
A coding error	Caught on the pallet	Impossible by construction
Marbete reconciliation	By hand, at month end	Automatic, per order
A stamp discrepancy	Visible weeks later	Visible the same day
The coder and the ERP	—	Neither is replaced

From two keying entries per bottling job to zero. From manual, after-the-fact stamp reconciliation to automatic reconciliation per order. From coding errors caught on the pallet to coding errors made impossible by construction.

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 5-10 months.
• The bulk of the value is not typing time: it is the coding incidents avoided, which in premium product and contract production carry reputational as well as economic cost.
• Marbete consumption reconciled per bottling order instead of after the fact.
• And a bottling supervisor who no longer learns a new screen, because the order travels on its own.

Estimated payback 5-10 months. The bulk of the value is not typing time but the coding incidents avoided, which in premium product and contract production carry reputational as well as economic cost. *Estimate to be validated.*

And the fair question from the production manager

"Isn't a classic integration project cleaner?" — it is, the day it reaches the top of the IT list, and that is the problem: this is the classic case of a modern coder still isolated because there was never time. Stitching through the existing APIs replaces neither system and needs no shutdown window. Where a field still has to be read from a printed order, that is an anchored task, where the best models drop below 1.5% error [1], and the supervisor confirms the job before the first bottle is coded. A discrepancy between order and coder is shown before the job starts, not after the pallet is wrapped.

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