From the grower's text message to action, in seconds

The problem

The harvest plan changes by text message, and it reaches one phone.

During harvest, dozens of associated growers send changes through informal channels that reach one person, not the planning system.

• During harvest, associated growers send changes all day long: a plot picked a day early because rain is coming, a smaller load than announced, a variety arriving before its turn.
• Those messages land on one person's phone or mailbox, mixed with everything else, not in the planning system that reserves tanks and sets the scale schedule.
• When that person reads the message late, the tank that should have taken that load is already reserved for another one.
• The result is confusion at the scale line, trailers waiting in a queue and grapes sitting in the sun while someone reshuffles tanks by hand and phones three growers back.

How it fits the IRIS system

Connect in chaotic-sources mode — mailbox and messaging in silent copy, twenty-four seven.

Connect chaotic-sources mode — mailbox/WhatsApp Business in 24/7 silent copy. The LLM extracts intent, entities (grower, plot, variety, date) and matches against the ERP, returning the precise action to the right role.

Reading the message sooner is only half the job. What actually helps the cellar is getting it back already matched against the harvest plan: which tank changes, which load moves, who has to act. That is what comes back, to the right role and in seconds. And everything that came in stays on record, so the end-of-campaign settlement with each grower is not argued from memory.

See the full IRIS architecture →

Before and after

The alert on one phone versus the alert matched to the plan

Aspect	Today	With iLEAN Connect
When the grower's change is read	When its recipient gets a moment	In seconds
Tank reservation	Already taken by another load	Adjusted before the trailer arrives
Scale line during peak	Hours of confusion	Loads arrive where expected
Which grower, plot and variety	Read between the lines	Extracted and matched with the ERP
Who learns about it	Whoever owns the phone	The role that has to act
Final call on the tank plan	The recipient	The cellar master — iLEAN proposes

hours of scale-line confusion → seconds; rescheduling based on the real alert, not on what got read late.

Impact estimate

Impact estimate per harvest — 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 the number of associated growers and how often they send changes.
• Hours of scale-line confusion reduced to seconds of reaction, in the weeks when every hour counts. The cellar master stops being the bottleneck through which every change has to pass.
• Rescheduling made on the real alert, not on whatever got read late.
• And fewer loads waiting in the sun while tanks are reshuffled by hand, which protects the fruit as much as the schedule. Grapes that wait less at the scale arrive at the tank in better condition.

estimated payback of 3-7 months depending on the number of associated growers and alert frequency. Estimate to validate.

And the fair question from the production manager

"Growers write however they like, with abbreviations and voice notes — will it understand them?" — extracting intent, grower, plot, variety and date from a short message is an anchored task, where the best models drop below 1.5% error [1]. And the match against the ERP is a second net: a plot that does not belong to that grower, or a variety it does not grow, is flagged instead of acted on.

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