In continuous process, a supply delay is a countdown

The problem

The alert lands in one inbox, and the train keeps consuming.

The critical alert arrives through a human channel: an email from the paraxylene or MEG supplier, a message from the marine terminal agent, a call from the carrier, a specification change from the catalyst supplier. It lands with one specific person, who cannot keep an inbox open in real time because they have another job to do. By the time they read it, the window to react at no cost has already narrowed. And in an integrated complex the effect propagates: a feedstock problem in the first unit becomes, three links downstream, an empty product silo and an unserved customer. The email read late in an office gets paid for at the loading dock two weeks later.

• The critical alert arrives through a human channel: an email from the paraxylene or MEG supplier, a message from the marine terminal agent, a call from the carrier, a specification change from the catalyst supplier.
• It lands with one specific person, who cannot keep an inbox open in real time because they have another job to do. By the time they read it, the window to react at no cost has narrowed.
• In an integrated complex the effect propagates: a feedstock problem in the first unit becomes, three links downstream, an empty chip silo and an unserved customer.
• An orderly rate reduction costs a fraction of running out of feedstock with the train hot. The difference between the two is decided by how early somebody knows.

How it fits the IRIS system

Connect in chaotic-sources mode — silent copy, twenty-four seven, with the numbers attached.

Connect chaotic-sources mode.

Reading the alert sooner is not the value. The value is the cross-check: how many hours of coverage remain at current consumption, and by when the rate decision has to be made. That comes back to the production manager as a number with a deadline, not as a forwarded email — and the decision stays theirs. Reducing rate on a Tuesday afternoon is a plan; running dry on Thursday night is a restart.

• A dedicated mailbox and a messaging number on silent copy, twenty-four seven.
• The model extracts intent — delay, quantity change, non-conformance, specification change, berthing window — and entities: supplier, material, volume, deadline.
• Cross-reference against live data: tank inventory, hourly consumption, open orders.
• It returns a concrete action with its time horizon to the relevant role, not a forwarded email: how many hours of coverage remain at current consumption and by when the decision has to be made. The decision stays with the person; what changes is that they decide with the number in front of them.

See the full IRIS architecture →

Before and after

The alert in the inbox versus the alert cross-checked

Aspect	Today, in the inbox	With iLEAN Connect
When it is read	When the recipient opens their email	Second zero
What you receive	A supplier's message	Hours of coverage and a decision deadline
Who it reaches	Whoever it was addressed to	The role that must decide
Feedstock shortage	A trip with the train hot	An orderly rate reduction, in time
Berthing window change	Found out at the terminal	Cross-checked against tank level
Who decides	The recipient, late	The manager, with the number in front of them

from hours of human lag to seconds. From reacting to a fait accompli to deciding with margin on current data. From a trip caused by feedstock shortage to an orderly rate reduction.

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 3-7 months, depending on the frequency of supply incidents.
• Unplanned shutdowns from supply disruption avoided — the most expensive outcome a late email can have.
• Carrier and vessel demurrage not paid, because the berthing window is known when it changes.
• From hours of human lag to seconds, and from reacting to a fait accompli to deciding with margin.

estimated payback 3-7 months depending on the frequency of supply incidents; measured in unplanned shutdowns from supply disruption avoided and carrier demurrage not paid. *Estimate to validate*.

And the fair question from the production manager

«Are you going to read my planners' email?» — no. iLEAN sits on silent copy of a dedicated mailbox and a company messaging number: it sees only what suppliers, terminal agents and carriers send to those operational channels. Classifying the intent of an alert — delay, quantity change, specification change, berthing window — is an anchored task where the best models drop below 1.5% error [1], and the rate decision is taken by a person. If a message is ambiguous — a delay with no new date — it is routed as uncertain, with the original attached, rather than turned into a number that does not exist.

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