WhatsApp from the orchard that replans the line

The problem

The orchard knows at seven in the morning. The line finds out in the afternoon.

Critical fruit-availability alerts -harvest delay, lower yield, weather-driven quality change- arrive by WhatsApp or email to specific people who can't watch their phone in real time while on the plant floor. By the time they're read, a line is already planned to start that afternoon with a volume that won't arrive, causing hours of downtime or last-minute reassigned staff.

• Avocado harvest depends on weather and on cutting crews: rain stops the picking, a crew does not show, an orchard yields less than estimated or the fruit comes in with a different quality.
• Growers report it the way they always have — a WhatsApp to the buyer, a voice note, an email — to specific people who cannot read their phone while they are on the floor.
• By the time it is read, the pulping or packing line is already planned to start that afternoon with fruit that will not arrive.
• The result is hours of idle line, or staff reassigned at the last minute, in the few months when almost the whole year's volume goes through the plant.

How it fits the IRIS system

Connect in chaotic-sources mode — silent copy on the grower channel, day and night.

Connect chaotic-sources mode: iLEAN sits in 24/7 silent copy on WhatsApp Business or a dedicated mailbox. An LLM extracts intent (harvest delay or volume change), entities (orchard, date, estimated volume) and cross-checks it against the active receiving plan in the ERP, returning the plant manager a concrete action, not just a notification that a message arrived.

The message already contained everything: which orchard, how late, how much less fruit. What was missing was someone reading it the moment it arrived and knowing which line depended on that truck. That is what comes back — the action, not the notification. In a season this short, that difference is counted in line hours.

See the full IRIS architecture →

Before and after

The grower's message today versus the message crossed with the plan

Aspect	Today	With iLEAN Connect
When the delay is known	When someone checks their phone	The moment it is sent
What arrives	A message among a hundred	Orchard, new arrival time and expected volume
What it is compared against	The memory of whoever reads it	The active receiving plan in the ERP
The afternoon line start	Planned on fruit that will not come	Replanned before the crew is called
Voice notes from the orchard	Listened to later, if at all	Transcribed and parsed
Who approves the change	Whoever happens to see it	The plant manager, with the action proposed

2-3 hours of human reaction latency → seconds. Replanning based on real-time data, not on what was read late.

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 critical alerts arrive during the season.
• The return is avoided unplanned downtime: lines that stop waiting for fruit that was never going to come.
• Human reaction latency drops from 2-3 hours to seconds, which in practice means the change is made before the afternoon shift is called in.
• And replanning is done on what the orchard said this morning, not on what someone read too late.

Estimated payback 3-7 months depending on how often critical alerts arrive in season, from avoided unplanned downtime. *Estimate to be validated.*

And the fair question from the production manager

“Growers write however they like, with voice notes and abbreviations — can it understand them?” — pulling the intent (delay, lower volume, quality change) and the entities (orchard, time, volume) out of a short message is an anchored task, where the best models drop below 1.5% error [1]. And iLEAN only proposes: the plant manager confirms the replanning before any line changes, and every accepted or rejected proposal is kept as history for the next season.

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