Two plants in the group, finally stitched together

The problem

Two plants in the group, connected by phone calls and emails.

Each drum's spring-latch closure is made on a fairly modern press at the group's sister plant, which already has its own API but is isolated from the assembly plant's ERP. Today, when a new order comes in, someone calls or emails to say how many closures of which format are needed and by when — a message that, if lost or late, leaves the assembly line stalled waiting for closures.

• Each drum's spring-latch closure is made on a fairly modern press at the group's sister plant. The closure has to match the drum diameter and the rating of each order.
• That press already has its own API, but it is isolated from the assembly plant's ERP. It is modern enough to integrate; it simply never got to the top of the IT list.
• When a new order comes in, someone calls or emails to say how many closures of which format are needed and by when. The same data already exists in the assembly ERP, typed once more by hand.
• If that message is lost or late, the assembly line stalls waiting for closures — or the wrong format arrives for the drum diameter on the line. Either way, finished drum bodies wait on the floor for a part made somewhere else. The line is ready; the closures are not.

How it fits the IRIS system

Connect stitches the closure press to the assembly ERP — APIs, not phone calls.

With iLEAN, when the assembly ERP flags an active order, it automatically publishes the required closure format and quantity to the sister plant's press via API. An order change becomes an automatic update, not a phone call.

Neither system is replaced and nobody has to retype an order. When the assembly ERP activates an order, the press at the other plant already knows the format and quantity, and an order change travels the same way instead of waiting for someone to remember to call. The phone call disappears; the decision at each plant stays with its people.

See the full IRIS architecture →

Before and after

Closure planning between plants today versus stitched by API

Aspect	Today	With iLEAN Connect
How the press learns of an order	A phone call or an email	Published from the ERP
Closure format for each drum diameter	Repeated by hand	Sent with the order
An order change	Another call, if someone remembers	An automatic update
Assembly waiting for closures	Recurring stoppages	Planned between plants
Wrong closure format delivered	Found at assembly	Prevented at the source
Press and ERP	—	Both kept, neither replaced

Before: closure planning by phone or email, with line stoppages from mismatches. After: automatic planning between plants.

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 fewer assembly-line stoppages due to closure shortages. Each stoppage avoided is a shift of assembly that delivers instead of waiting.
• Closure planning moves from phone and email to automatic planning between the two plants of the group.
• Order changes stop depending on a message reaching the right person at the other site.
• And the format sent to the press is the format in the order, which removes a whole class of mismatches. The two plants of the group plan from the same order instead of from two versions of it.

Estimated payback 3-6 months, from reduced assembly-line stoppages due to closure shortages. Estimate to be validated.

And the fair question from the production manager

“What if the ERP order is wrong and the press makes the wrong closures?” — the press receives exactly what the order says, so a wrong order would be wrong either way; the difference is that now it is visible in one place. Mapping an order to closure format and quantity is an anchored task, where the best models drop below 1.5% error [1], and the planner confirms new mappings before they go live. Once confirmed, the same order always produces the same closure request.

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