Voice on the Weighing Floor

The problem

The supervisor knows what went wrong. The shift report does not.

The production supervisor walks the floor all day between weighing, mixing and packaging with hands full. Critical information about raw materials or operators dies at shift end because they can't type.

• The production supervisor spends the shift walking between the weighing booths, the ribbon blender and the packing line, gloves on and often a sample bag in hand.
• They see what no system captures: a protein lot that cakes in the hopper, a fiber that bridges in the dosing screw, an operator who needs another run through the vitamin premix procedure.
• Typing it means walking back to an office through a hygiene-zoned plant, so it is told verbally at handover or written on a notebook page that nobody reads again.
• By the next shift most of it is gone, and the same caking or bridging is diagnosed from scratch the next time that raw material comes in, by a different person, with no memory of the last time.

How it fits the IRIS system

Connect in voice mode — the supervisor speaks, iLEAN files it against the formula.

Connect voice mode: Bluetooth headset and a wake word. iLEAN listens, structures the note linked to the active formula/shift, saves it, and replies by voice or push in seconds.

Only what the supervisor decides to say is recorded, when they decide to say it, after a wake word. There is no open microphone on the floor. That is what keeps the tool in use after the first month and stops it being read as a way of tracking people. In a blending plant the supervisor is the person who connects raw material behavior with what happens in the blender; giving that knowledge somewhere to go is the whole point.

See the full IRIS architecture →

Before and after

The supervisor's round today versus a round with a headset

Aspect	Today	With iLEAN Connect
Note on how a raw material behaves	Told at handover, then lost	Stored against the lot and the formula
Caking or bridging that keeps coming back	Rediagnosed each time	Visible as a pattern with dates
Effort to record a floor observation	Walk to the office and degown	Speak while walking
Reply to a quick question on a batch	Phone call or nothing	Voice or push in seconds
Handover between blending shifts	A few verbal minutes	Structured summary ready
Operators' conversations on the floor	—	Never recorded

Knowledge that dies at shift end → enters the system in real time and gets cross-referenced with live data.

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.

• No standalone payback: this is an enabling piece. Its value is strategic and we do not attach a month count to it, because inventing one would not hold up.
• What it changes is that knowledge that dies at shift end enters the system in real time and gets cross-referenced with live formula and batch data.
• It feeds the evidence pack and the allergen flagship: without the supervisor's notes, a batch dossier has gaps exactly where the floor knows most and writes least.
• And continuous improvement starts working on real plant data instead of the memory of whoever happened to be on shift that week. A recurring raw material problem finally gets enough dated evidence to justify a conversation with the supplier.

Strategic value, not directly monetizable · enables continuous improvement with real plant data

And the fair question from the production manager

"Is the plant going to record my operators?" — no. Only the supervisor wears the headset, it activates with a wake word and stores only what is dictated. Turning a spoken note into area, raw material, formula and severity is an anchored task with your own vocabulary, where the best models drop below 1.5% error [1], and anything unclear is sent back for confirmation instead of guessed.

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