The mill alert that arrives on time

The problem

The message that breaks the week lands in a mailbox nobody is looking at.

The mill emails or messages about a delayed heat, or a customer asks to change a measurement already scheduled. The message sits in an inbox nobody checks in real time while handling the counter or the phone, and by the time it's read, decisions already made fall apart.

• The mill writes to say a heat is running late. A customer messages to change a size already scheduled for tomorrow's cutting list. Both arrive in plain language, at a personal address, at the worst possible hour.
• Whoever they were sent to is on the counter or on the phone selling. The message sits unread for two or three hours, competing with a hundred others that matter far less.
• By the time it is read, the promise has already been made: material committed at the counter, a date given to a workshop, a cutting list sequenced around a heat that is not coming.
• It is not solved by asking people to check their inbox more often, because the counter does not stop.

How it fits the IRIS system

Connect in chaotic-sources mode — mailbox and business number, in silent copy around the clock.

Connect chaotic-sources mode: mailbox or WhatsApp Business in silent copy 24/7. The LLM extracts intent and entities, cross-checks active ERP orders, and returns the specific action to the manager.

Reading the alert sooner is only half of it. What comes back is the cross-check: which open orders depend on that heat, whether another format in stock covers the same grade and thickness, and which customer has to be called before they send a truck for material that is not there.

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 it is read	2-3 hours later	Seconds
A delayed heat from the mill	Found out with the order already promised	An alternative from stock proposed
A size change from a customer	Found out after the cut	Caught before the batch starts
Rescheduling against	This morning's picture	Live orders and live stock
What arrives	One more email	Which order to move and whom to call
Who decides	Whoever opens the mailbox	The manager — iLEAN proposes

2-3 hours of human latency → seconds, with rescheduling based on real-time data.

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 3-7 months, depending on how often a critical alert actually lands.
• Measured in promises kept: material committed at the counter against a heat that was never going to arrive is the expensive kind of mistake.
• From 2-3 hours of human latency to seconds, with the cutting list resequenced on live data rather than on this morning's picture.
• And a size change caught before the batch starts instead of after the sheet is cut, which is the difference between a correction and a scrapped piece.

Estimated payback of 3 to 7 months depending on critical alert frequency. Estimate to be validated.

And the fair question from the production manager

“Are you going to read my team's email?” — no. iLEAN sits in copy on a dedicated mailbox and a business messaging number: it sees only what mills and customers send to those commercial channels. On the extraction side, classifying intent — a delay, a size change, a claim, a routine acknowledgment — and pulling out order, heat, grade and thickness is an anchored task where the best models drop below 1.5% error [1], and the action itself is always taken by a person.

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