Planning while walking the bay, hands free

The problem

The real state of the plant is known by one person, walking.

The plant's most valuable operational knowledge is generated while walking, and it is exactly the knowledge nobody records. With foundry noise and both hands busy, typing is not a realistic option: so the real state of the plant is passed on by phone at handover, and much of it is lost on the way. The cost never shows up as a line item, but it shows up every morning: the incoming shift reconstructs what was already known, repeats checks and makes decisions without the context somebody had twelve hours earlier. Much of the schedule slippage is born right there.

• How many molds the line is really producing, how many raw containers are piling up in front of machining and what can be promised to the customer this afternoon is learned on foot, not on a screen.
• With foundry noise and both hands busy, typing is not an option. So it goes into a notebook and is passed on by phone at handover — and much of it is lost on the way.
• The incoming shift reconstructs what was already known twelve hours earlier, repeats checks and decides without the context somebody had. The same container gets counted three times a day.
• A good part of the schedule slippage is born right there, and it never appears as a line item because nobody measures what was forgotten.

How it fits the IRIS system

Connect in voice mode — the planner speaks, iLEAN structures and answers.

Connect in voice mode with an industrial Bluetooth earpiece:

The earpiece does two things at once. It records the observation the moment it happens, anchored to that furnace, that pattern, those containers. And it answers out loud with the data already crossed, so the question asked at the molding line — is there coverage for this part number tomorrow? — is resolved before the planner reaches the next bay. At the end of the shift, the handover is already written, resource by resource, and the incoming planner starts from the real state instead of reconstructing it. Humans in command taken literally: it records what the planner chooses to say, when they choose to say it, and nothing else.

• Recognition robust to industrial noise, trained on the language of the trade: heat, riser, core, pattern plate, nodularity, flash.
• iLEAN listens, structures the observation and anchors it to the actual resource: that furnace, that pattern, those containers.
• It answers out loud in seconds, with the data already cross-checked: "with those three containers that part number has two days of coverage, shall I propose it for tomorrow's schedule?".
• At the end of the shift, handover receives the real state in writing instead of a phone call.

See the full IRIS architecture →

Before and after

Today's walk versus the captured walk

Aspect	Today	With iLEAN Connect
Observation at the line	A notebook, if that	Recorded and anchored to the resource
Raw coverage by part number	Counted by eye, kept in the head	Answered out loud in seconds
Shift handover	A phone call from memory	The real state, in writing
Operational decisions	No trace	History searchable by furnace, pattern or container
Planner's time	Walk, then type	Speak while walking
Operators on the floor	—	Not recorded

from knowledge that dies at the end of the shift to a continuous searchable record · from handover by phone and memory to handover with the real state written down · from no traceability of operational decisions to a history searchable by resource.

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 4-9 months.
• 45-90 minutes a day recovered for the role that walks the plant, today spent walking back to a desk to type what was seen.
• A direct improvement in handover quality between shifts: the real state written down instead of a phone call.
• And a searchable history of operational decisions, by resource, that today does not exist anywhere.

estimated payback 4 to 9 months, with 45-90 minutes a day recovered for the role that walks the plant and a direct improvement in handover quality between shifts. *Estimate to validate*.

And the fair question from the production manager

“Will it understand a melter's vocabulary over the noise of the molding line?” — recognition is trained on the language of the trade — heat, riser, core, pattern plate, nodularity, flash — and prepared for industrial noise, and the headset is compatible with hearing protection. Structuring a dictated observation into resource, symptom and quantity is an anchored task where the best models drop below 1.5% error [1], and what was understood is read back so the planner can correct it on the spot, before it is stored.

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