One mistyped digit cannot be allowed to cost a recall

The problem

A thirty-second operation, and a recall if it goes wrong.

Manual typing on the coder panel is still the standard method, not because integration is impossible, but because "integrating it properly" has never been a priority against producing. The risk is asymmetric: the operation costs thirty seconds and its failure costs a recall. One wrong digit and half a production run goes out miscoded. Caught, it is rework; missed, it is a recall. And with a portfolio of hundreds of shades and several destination markets, the frequency of change — and therefore of typing — is enormous.

• In a kit the coder does not print one datum: it prints three — tube, developer bottle and carton — which also have to agree with each other.
• Typing on its panel is still the standard method, not because integration is impossible, but because it has never been a priority against producing.
• The risk is asymmetric: the operation costs thirty seconds and its failure costs a recall. Caught, it is rework; missed, it is a recall.
• With hundreds of shades and several destination markets, the frequency of changeovers — and therefore of typing — is very high.

How it fits the IRIS system

Connect stitches two modern systems through APIs, replacing neither.

Connect stitches two modern systems through their APIs, replacing neither. When the ERP marks the order as started, it automatically publishes to the coder the lot, the expiry date and the data set matching the destination market. The shade change stops being a keystroke on the floor and becomes a state change in the ERP. And the coder returns confirmation of what it actually printed, stored per batch as evidence.

The machine returns confirmation of what it actually printed, not of what it was asked to print. That difference is what turns the coding record into evidence instead of an assumption filed next to a retained sample.

See the full IRIS architecture →

Before and after

The isolated coder versus the stitched coder

Aspect	Today	With iLEAN Connect
Keystrokes per shade changeover	One per kit element	Zero
Tube, bottle and carton agreeing	Depends on typing the same thing three times	Guaranteed from the order
Coding errors per year	Several, some with recall risk	Zero of human origin
What was actually printed	Rebuilt from a retained sample	Automatic per-batch record confirmed by the machine
Destination market data	Typed	Published from the live order
Either system	—	Neither is replaced

Impact estimate

Estimated impact — to validate 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.

• Estimated payback 3-6 months — among the fastest, because the machine already ships an API and the integration effort is low.
• A sharp reduction in recall risk from wrong coding, which is the most frequent failure mode in this product family.
• From several typing errors a year — some with rework — to zero of human origin.
• An automatic per-batch record of what the machine confirmed it printed.

fast payback, estimated 3 to 6 months, and a drastic reduction in recall risk from miscoding, which is the most frequent failure mode in this product family. *Estimate to validate*.

And the fair question from the production manager

«Isn't a classic integration cleaner?» — it is, once the project reaches the top of the list, and that is exactly the problem: it has not for years, because there is always something more urgent than producing. This stitching replaces neither system and does not touch the coder's program, so it competes for neither the shutdown window nor a large project budget.

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