The schedule travels by EDI; the problems do not

The problem

What breaks the week never travels on the formal channel.

The plant runs against a sequenced schedule with no inventory buffer. A mix change communicated by email and never folded into the plan means, two days later, either a line stopped for lack of coil in the right gauge, or an urgent shipment paid at urgent prices. That cost never gets booked as "an information problem". It is booked as premium freight, as overtime or as a customer penalty, which is exactly why nobody attacks it at source. The same pattern repeats on inbound supply: the steel supplier flags a delay or a heat deviation, and that information travels through a channel no system reads.

• The formal schedule travels by electronic data interchange. The sequencing mix change, the coil delay and the truck stuck at the border travel by email or messaging.
• All of it lands in one person's inbox and gets read when there is no margin left to replan.
• Two days later the result is a line stopped for lack of coil in the right gauge, or an urgent shipment paid at urgent prices.
• The cost is never booked as "an information problem". It is booked as special freight, overtime or a customer penalty. That is why nobody attacks it at source.
• The pattern repeats upstream: the steel supplier flags a heat deviation through a channel no system reads.

How it fits the IRIS system

Connect in chaotic-sources mode — humans in command, always.

Connect in chaotic-sources mode.

iLEAN does not replan on its own. It proposes the concrete action with its alternatives, and the planner decides and executes. The difference between an actionable alert and an email summary is what separates a useful tool from one more notification nobody reads.

• iLEAN sits on copy twenty-four seven on a dedicated mailbox and a business messaging number.
• The moment a message arrives it extracts intent: pull-ahead, delay, quantity change, material deviation or nonconformity.
• It matches that intent against live business system data: delivery schedule, coil coverage by gauge and grade, open orders.
• It issues an actionable alert to the planner and the logistics lead — not a summary of the email, but the concrete action with its alternatives.
• Humans in command: iLEAN proposes, the person decides and executes.

See the full IRIS architecture →

Before and after

Today's alert versus the cross-checked alert

Aspect	Today, in the inbox	With iLEAN Connect
When the message is read	When that person opens their email	The moment it lands, twenty-four seven
What you know when reading it	What the message says	The message crossed with coil coverage and open orders
What you receive	One more email	The concrete action and its alternatives
When the impact is detected	The next day	In minutes
If that person is on holiday	The alert waits	The alert goes out anyway
Who decides	The planner	The planner — iLEAN proposes

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 4-8 months, measured against lines already in your accounts.
• Urgent transport and reactive overtime you stop paying.
• Avoided risk of stopping the OEM's line because a supply issue was seen too late.
• From reactive replanning to preventive replanning, which is where today's margin leaks.

estimated payback 4-8 months, from the urgent transport and reactive overtime you stop paying, plus the avoided risk of stopping the customer's line because a supply issue was seen too late. *Estimate to validate.*

And the fair question from the production manager

«Are you going to read my people's email?» — no. iLEAN sits in copy on a dedicated mailbox and a company messaging number: it only sees what is sent to that address, which is exactly the channel these alerts arrive on today. On extraction reliability, classifying the intent of an operational message is an anchored task and the best models drop below 1.5% error [1] — and even then the alert goes to a person who decides.

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