The person who knows the plant best is the one who can log the least

The problem

The person who knows the kitchen best is the one who cannot stop to write.

The operational knowledge of a production workshop lives in the heads of two or three people with many years in the house. That the béchamel of one recipe runs thinner since the supplier changed. That the breading machine leaves bare pieces above a certain speed. That the new operator on the former needs support this week. None of that enters any system, and not for lack of willingness: asking that person to stop at a computer is asking them to stop doing their job. The result is that continuous improvement meetings run on memory rather than on history. And any improvement plan carries that limitation from day one.

• The operational knowledge of the kitchen lives in the heads of two or three people with many years in the house: that the béchamel of one recipe runs thinner since the dairy changed, that the breading machine leaves bare pieces above a certain speed, that the new operator on the forming machine needs support this week.
• None of that enters any system, and not for lack of willingness. The supervisor walks the shift with both hands busy — a tray, a probe, a bag of breadcrumbs — and asking them to stop at a computer is asking them to stop doing their job.
• Continuous improvement meetings run on memory rather than on history, and any improvement plan carries that limitation from day one.
• When one of those people is on leave or retires, the recipe knowledge that was never written down leaves with them.

How it fits the IRIS system

Connect in voice mode — iLEAN listens to the supervisor, not to the operators.

Connect in voice mode.

Humans in command taken literally: the earpiece records what the supervisor chooses to dictate, when they choose to, and answers through the same earpiece. That distinction is what keeps the case alive in month six instead of being perceived as monitoring.

• A Bluetooth earpiece with its own wake word.
• The supervisor dictates in their own words, with no format, without quoting batch numbers.
• The model structures the dictation as an incident, process observation or request, and ties it automatically to the active batch, line and shift.
• It returns confirmation and support through the same earpiece, so the conversation runs both ways without them taking their hands off production.
• Everything captured passes through early human validation before entering the master.

See the full IRIS architecture →

Before and after

Today's shift versus the captured shift

Aspect	Today	With iLEAN Connect
What the supervisor observes	Lost at the end of the shift	A structured record tied to the batch
A béchamel that runs thin since the supplier change	An anecdote in a meeting	An incident with a recipe and a date
Support for a new operator	Whoever is next to them	A request logged on the forming machine
Supervisor's effort	Stop at a computer	Speak while walking
Improvement meeting	Based on memory	Based on the incident history by reference and line
Ordinary kitchen conversation	—	Not recorded

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, and the case analysis says so plainly: this is an enabling piece whose return is indirect.
• Without this capture, continuous improvement keeps depending on two people's memory, and every other case in the matrix inherits that gap.
• What it enables is measurable elsewhere: recipe deviations caught earlier, a defect history by line that the coating vision case can lean on, onboarding that stops depending on who happens to be next to the new operator.
• The estimate is agreed with the production manager, against the incidents they know are repeated every month and never written down.

this is not a direct payback case and it is worth saying so plainly: the return is indirect and enabling. Without this capture, continuous improvement keeps depending on two people's memory. *Estimate to be validated with the production manager*.

And the fair question from the production manager

“Is this for monitoring my team?” — no: iLEAN listens to the supervisor, not to the operators, and only to what they choose to dictate after the wake word. There is no continuous listening and no tracking of where anyone is. On the technical side, turning a dictation into an incident, observation or request tied to the active batch is an anchored task, where the best models drop below 1.5% error [1], and every record passes through the line-side tablet before entering the master.

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