The earpiece that returns the shift’s knowledge to the system

The problem

The shift's knowledge lives in the shift lead's head and dies when the shift ends.

In a dairy processing plant, the shift lead is not sitting in front of a screen: they are walking, hands busy, between the slicer, the packing machine and the cold room, for seven or eight hours straight. They are the person who first sees, first smells and first knows when something starts going wrong:

• Scrap at the slicer — a batch starting to give more trim than usual, long before the data shows up in any end-of-shift report.
• Reinforcement needed at a station — an operator falling behind at packing who needs support before the backlog piles up.
• A temperature drift in the cold room — something they notice walking past, with nobody having asked and nowhere to note it at that moment.

Nobody asks them anything because they cannot type while walking. That critical knowledge stays in their head until the end of the shift — and often does not even make it there: it dilutes between the shift change, the handover and the next urgency. The continuous improvement plans then depend on a weekly meeting where someone tries to remember what happened, with no data, from memory.

How it fits the IRIS system

Connect in earpiece mode — the shift lead takes notes by voice, without freeing their hands.

The earpiece is not a system that listens to the shift lead all day: it is a tool that the shift lead activates when they want to note something down. They decide when they speak and what they say. Connect structures it, links it to the plant's active context and returns it in seconds.

The shift lead says "iLEAN, note this". They dictate the incident in their own words, while walking. Connect structures it and links it to the active SKU, batch, line and shift. Seconds later, the answer arrives by voice to the earpiece or as a notice on the tablet.

How Connect operates in earpiece mode in a dairy processing plant:

• A custom wake word — the shift lead opens the dictation with their own phrase, "iLEAN, note this", which only they say deliberately. Outside that phrase, the earpiece neither transcribes nor sends anything.
• Free dictation, no forms — they speak in their own words, just as they would tell a colleague: "batch 4 on the slicer giving more trim than normal for the last half hour".
• The LLM structures the incident — Connect turns the dictation into an incident with clear fields, automatically linked to the SKU, batch, line and shift active at that moment, without the shift lead having to say or remember that data.
• A voice reply in seconds — the answer arrives at the same earpiece or, if more context is needed, as a notice on the nearest tablet: "quality confirms at 12:15, want me to remind you?". The shift lead keeps walking, the incident already recorded.
• A history available for continuous improvement — every dictated incident stays linked to its batch and its line, available for whoever needs it later, instead of being lost when the shift ends.

See the full IRIS architecture →

Before and after

Knowledge in the shift lead's head vs. knowledge in the system

Aspect	Knowledge in someone's head	Knowledge in the system with Connect
An incident spotted while walking	Stays in the head until the shift ends, if it gets there	Dictated on the spot, hands free
Link to SKU, batch, line and shift	Depends on someone remembering it later	Automatic, the system already knows what is active
Support or confirmation	The shift lead physically looks for someone	A voice reply on the earpiece in seconds
The shift-change meeting	A memory exercise about what happened	A review of incidents already recorded
Continuous improvement plan (WCM/TPM)	Fed by recollections at the weekly meeting	Fed by real data captured on the floor
The incident's record	Does not exist, or only if someone wrote it by hand	Linked to batch and line, available afterwards

Impact estimate

Impact estimate for your plant — to be validated with your own 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.

• Dairy processing plant with shift leads walking most of the day between slicing, packing and cold rooms, hands busy and with no practical access to a keyboard.
• Connect pilot in earpiece mode on one shift or one line — without changing the shift lead's role, without adding administrative tasks.
• The return is not an accounting saving of hours: it is enabling WCM, TPM and Lean with real data instead of memory reconstructions at the weekly meeting.
• The impact estimate on detected incidents and on the quality of the improvement plans is validated with the plant's continuous improvement team, not invented here as a generic figure.

And the fair question from the plant manager

"Doesn't this end up being a system that monitors what my shift lead says?" — no. It is exactly the other way around: the earpiece hears nothing until the shift lead themselves activates it with their wake word. There is no background recording, no continuous transcription, nobody listening to their conversation with an operator. The only thing entering the system is what they themselves decide to dictate as an incident — because they want it on the record, not because someone above is asking. The tool works for them, not on them.

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