The agave truck alert, read at second zero

The problem

The alert that changes the day lands on a phone nobody can look at.

The critical alerts in this sub-sector — a delayed or canceled agave truck, actual tonnage differing from what was committed, a change of field of origin, a price variation, a carrier incident, a notice from the certifying body about an inspection — arrive through informal channels into personal inboxes and phones. Nobody can watch their phone and close a batch at the same time. And in a plant that cooks in batches, reacting an hour late means a wasted autoclave load and an idle fermentation tank. The cost is not theoretical: it is a load that was not made and tank hours that do not come back. Worse, reacting late forces emergency purchases at worse prices, on the input that is already the plant's largest variable cost line.

• A delayed or canceled agave truck, actual tonnage below what was committed, a change of field of origin, a price variation, a carrier incident, a notice from the certifying body about an inspection: all of it arrives through informal channels into personal inboxes and phones.
• Nobody can watch a phone and close a batch at the same time. By the time the message is read, the autoclave is scheduled and the tank reserved. The person it was sent to is usually the one with the least time to read it.
• In a plant that cooks in batches, reacting an hour late is a wasted autoclave load and an idle fermentation tank. That cost is not theoretical: it is a load that was not made and tank hours that do not come back.
• And reacting late forces emergency agave purchases at worse prices, on the input that is already 60-70% of variable cost.

How it fits the IRIS system

Connect in chaotic-sources mode — silent copy on a corporate channel, twenty-four seven.

Connect in chaotic-sources mode. The flow:

iLEAN does not say you have a new message: it says what can be done about it. Which load moves up, which tank frees up and who needs to be told, with the information already joined up. The production manager still decides; what changes is that the alert arrives with its consequences worked out, and in time. In a batch plant, time is the only thing that turns an alert into a wasted load or into an ordinary Tuesday.

• iLEAN sits silently copied in 24/7 on a dedicated plant mailbox and a business messaging number. Never on anyone's personal phone: the channel is corporate and explicit.
• The moment a message lands, it extracts the intent — delay, tonnage change, origin change — and the entities: supplier, field, kilos, transport, date.
• It cross-references those against the cooking schedule and the live purchase orders in the ERP.
• It hands the production manager a concrete action: which load moves up, which tank frees up and who needs to be told.

Step 4 is where the difference lies. iLEAN does not say "you have a new message": it says what can be done about it. The decision stays human; what changes is that it arrives with the information already joined up, and in time.

See the full IRIS architecture →

Before and after

Today's alert versus the cross-referenced alert

Aspect	Today, on a phone	With iLEAN Connect
When the alert is read	When somebody looks at the phone	Second zero
Human lag	2-4 hours	Seconds
An autoclave load without agave	Discovered at loading time	Rescheduled while the alert is warm
The reserved tank	Sits idle	Freed or reassigned in time
Emergency purchases	At whatever price the day allows	Fewer, and planned
Who decides	The production manager	The production manager — iLEAN proposes

From 2-4 hours of human lag to seconds. From replanning once the truck has already failed to show up, to replanning while the alert is still warm.

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 4-8 months.
• Recovered cooking loads and fermentation tank hours that today are lost to late reaction.
• Fewer emergency agave purchases at worse prices, on the plant's largest variable cost line.
• And a production manager who replans while the alert is still warm instead of once the truck has already failed to show up.

Estimated payback 4-8 months from recovered cooking loads and tank hours, plus fewer emergency purchases at worse prices. *Estimate to be validated.*

And the fair question from the production manager

"Is iLEAN going to read my team's personal messages?" — no. It sits in copy on a dedicated plant mailbox and a business messaging number, and it only sees what suppliers and carriers send to those corporate channels. On extraction, classifying the intent of a supply alert — delay, tonnage change, origin change — is an anchored task where the best models drop below 1.5% error [1], and the reshuffle is executed by a person. Suppliers and carriers keep writing exactly as they do today; nobody is asked to fill in a form.

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