What the mill supervisor notices no longer gets lost

The problem

The person who knows the mill best has the least time to write it down.

The role that accumulates the most tacit knowledge about the mill's real state also has the least time to type. That knowledge disappears shift after shift.

• The mill supervisor walks silos, conditioning, the roller mill floors, plansifters and bagging every shift, with hands busy and dust everywhere. On a multi-floor mill, a full round can take most of an hour.
• They notice what no sensor reports: a sifter vibrating differently, a bearing getting warm, a choked spout, a leak at a pneumatic conveying line. Small things, until they are not.
• That knowledge is lost by the end of the shift, because writing it down means stopping, finding a terminal and typing. On a milling floor, the nearest terminal is often two floors away, and the next problem is already waiting.
• So the small defect noticed on Monday becomes the corrective maintenance stop on Thursday. And the next supervisor starts the shift without knowing any of it.

How it fits the IRIS system

Connect in voice mode — the supervisor speaks, the system structures.

Bluetooth headset + Connect voice mode. The AI transcribes, classifies by category (maintenance, quality, safety) and routes the action to the right role without opening any app.

The supervisor does not open an app or fill in a form: they say what they see in their own words. The AI turns it into a classified, actionable entry and sends it to whoever has to act. Over months, the log shows which sifter or which spout keeps coming back, which no shift report ever made visible.

See the full IRIS architecture →

Before and after

The rounds today versus the rounds being heard

Aspect	Today	With iLEAN Connect
An observation on the sifter floor	Remembered, maybe	Logged in seconds
Classification	None	Maintenance, quality or safety
Who gets it	Whoever the supervisor meets	The responsible role, directly
Supervisor's hands	Need to stop and type	Stay free
Knowledge at shift change	Verbal, partial	A structured log
Recurring issues	Invisible	Visible as a pattern

From observations lost by the end of the shift, to a structured, actionable log in seconds.

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, from fewer corrective maintenance incidents caught too late. A stop on the roller mill bank costs tonnage that cannot be recovered later in the shift.
• Observations stop being lost at shift change and become a structured log in seconds. The next shift starts with a list of open points instead of a hurried verbal handover.
• Maintenance receives early warnings on roller mills and sifters it would otherwise only see as a stop. A warm bearing reported on Monday is a planned intervention, not a breakdown on Thursday.
• And the tacit knowledge of the mill's most experienced role starts to stay in the plant. When that supervisor retires or moves on, what they knew does not leave with them.

Estimated payback of 4-8 months from fewer corrective-maintenance incidents caught late. Estimate to validate.

And the fair question from the production manager

"With the noise of a milling floor, will it understand anything?" — the headset's microphone is directional and filters ambient noise, and turning a spoken observation into category, equipment and action is an anchored task, where the best models drop below 1.5% error [1]. When confidence is low, the system reads back what it understood and asks for confirmation. Nothing reaches maintenance or quality as an order until the classification is clear.

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