The line lead's knowledge no longer dies at shift's end

The problem

How the lines affect each other is only known by whoever walks between them.

Critical operational knowledge about how the lines behave relative to each other lives in the line lead's head and never enters the system in real time.

• The line lead spends the shift on foot: from the press to the dryer, from the glazing line to the kiln entry, with hands busy and no keyboard within reach.
• They are the person who knows that a pressure drift at press 2 shows up as edge chipping on the glazing line, or that a slow kiln entry backs up the boxes of fresh tiles waiting to be fired.
• That knowledge is passed on verbally at shift change, or in a few lines on the shift report. Most of it never enters any system, and nobody asks for it because nobody expects the line lead to stop and type.
• So continuous improvement works on the incidents that were written down, which are rarely the ones that explain how the lines interfere with each other. The rest leaves the plant at the end of every shift.

How it fits the IRIS system

Connect in voice mode — a custom wake word, and the incident is dictated walking.

Connect voice mode. Custom wake-word. A language model structures the dictation as an incident linked to the active line/shift. Feedback by voice or push.

Nobody asks the line lead to stop and type. They say the wake word, describe what they see, and keep walking. The incident lands tied to the line and the shift, next to the live data of the other lines, and the line lead gets support back in seconds. Over a month, those dictations become the map of how the lines really interfere with each other.

See the full IRIS architecture →

Before and after

The line lead's knowledge today versus knowledge captured by voice

Aspect	Today	With iLEAN Connect
Incidents seen between lines	Remembered until shift change	Dictated on the spot
Link to line and shift	Reconstructed afterwards	Automatic from context
Press drift seen at glazing	A comment in the handover	A record tied to both lines
Feedback to the line lead	None	By voice or push in seconds
Recurring cross-line problems	Invisible in the reports	Searchable by line and cause
Line lead's hands and attention	Busy	Still on the line

Knowledge dying at shift's end → entering the system in real time and cross-referenced with live data from the other lines.

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: we do not put a figure on it because its value is hard to monetize upfront, and we prefer to say so.
• It is an enabling piece: every real continuous-improvement project on a multi-line process depends on this information existing.
• Knowledge that died at shift's end enters the system in real time, cross-referenced with live data from the other lines.
• The value is best sized with your continuous-improvement manager, against the cross-line incidents you cannot explain today. The first weeks of dictations usually show which ones they are.

Hard to monetize upfront, but all real continuous improvement on a multi-line process depends on this information existing. *Estimate to validate* with the continuous-improvement manager.

And the fair question from the production manager

“Will it understand a dictated incident with kiln noise behind it?” — turning a short dictation into an incident with line, machine and symptom is an anchored task, because the lines, machines and defect names come from a closed list, and there the best models drop below 1.5% error [1]. And whatever the system understands is read back to the line lead by voice, so a wrong line or machine is corrected before it is stored. Nothing dictated reaches the ERP without passing the same human verification as the rest.

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