The paint superintendent in command — voice, hands-free, CMMS

The problem

Operational knowledge dies at the end of the shift — and the CMMS never finds out.

On an EMS plant's metal chassis paint line, the maintenance superintendent is not sitting in front of a screen: they are walking — pre-treatment, booths, application robots, curing oven, inspection table, spare parts warehouse — hands busy, all day. They cannot type while walking between machines, and nobody asks them anything because they are out of the office. They are the person who first sees what no sensor reports:

• Nozzle 3 is losing flow — a behavior change they notice in the spray cloud long before it shows as a defect on the inspection table.
• The sacrificial filter has 220 hours against the recommended 200 — a consumable-life data point that today lives in their head or, with luck, in a notebook.
• Robot A2 has a squeal — the early signal of a mechanical problem that has not yet triggered any formal alarm.

All that critical information about the equipment's state and life dies at the end of the shift or degrades to whatever gets remembered into a notebook. And the structural consequence is worse: the preventive and predictive plans run by calendar, not by real observation, because the observations persist structured in no system.

How it fits the IRIS system

Connect in voice mode with an earpiece — the superintendent dictates and the CMMS receives the structured order.

The earpiece is not a system recording the superintendent during the shift: it is a tool they activate voluntarily when they want to note something on their round of the paint line. They decide when to speak and what to dictate. Connect structures it, ties it to the plant's active context and creates the order in the CMMS with traceability.

The superintendent says "iLEAN, note this". They dictate the observation in their own words, without stopping walking. Connect structures it as a work order or incident tied to the active equipment, work order and shift, and creates it in the site's CMMS. Seconds later, the return arrives by voice on the earpiece or as a push to the leader's tablet.

How Connect operates in earpiece mode on a chassis paint line:

• A personalized wake-word — the superintendent opens the dictation with their own phrase ("iLEAN, note this") pronounced deliberately. Outside that phrase, the earpiece transcribes and sends nothing.
• Free dictation, no forms — they speak in their own words, just as they would tell their technician: "nozzle 3 in booth 1 is losing flow, have it checked before the color change".
• The LLM structures the order — Connect turns the dictation into a work order or incident with clear fields, automatically tied to the equipment, work order and shift active at that moment, without the superintendent reciting those data.
• Integration with the site's CMMS — the order is created directly in the system the plant already uses — SAP PM, FIIX or Infor EAM, among others — with full traceability of who dictated it, when and on which equipment.
• A return in seconds — the confirmation arrives by voice on the same earpiece or as a push to the leader's tablet: "order created, supplier confirmed, shall I tell you when it arrives?". The superintendent keeps walking, with the order already in the system.

See the full IRIS architecture →

Before and after

Knowledge that dies at the end of the shift vs. work orders created in the instant

Aspect	Knowledge dying at shift end	With Connect voice + CMMS
Operational knowledge captured	A residual fraction, whatever gets remembered into a notebook	Complete, structured and tied to equipment, order and shift
Work orders	Remembered at the end of the shift, if at all	Created in the CMMS at the instant of the observation
Consumable life (filters, nozzles)	Estimated from memory, changed late or too soon	Recorded by real hours and condition, machine by machine
MTBF / MTTR	Calculated on partial data	Calculated on real, rich data
Preventive / predictive plan	Triggered by calendar, blind to what is observed	Fed by the gemba's real observation
Shift handover	A rushed conversation and a notebook by the booth	The incoming shift inherits structured, dated orders

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.

• EMS plant with a metal chassis paint line — pre-treatment, booths, robots, curing oven — and a maintenance superintendent walking most of the day, hands busy and with no practical keyboard access.
• Connect voice pilot on the paint line — without changing the superintendent's role, without adding administrative tasks.
• The main return is not a countable hour saving: it is that the CMMS finally reflects the equipment's real state — every order created in the instant, tied and traceable — instead of the fraction surviving the end of the shift.
• There is an expected operational return too: the early detection of degradations — the nozzle losing flow, the filter past its hours, the robot's squeal — before they become a stoppage or a finish defect. That improvement is an estimate to be validated with the plant's maintenance manager, not a generic figure invented here.

And the fair question from the plant director

"Doesn't this end up being a system recording everything my superintendent says?" — no. It is exactly the other way round: the earpiece listens to nothing until the superintendent activates it with their wake-word. There is no background recording, no continuous transcription, nobody listening to their conversation with a painter or the oven technician. The only thing entering the system is what they decide to dictate — because they want it recorded, not because someone above asked. And the second doubt, "what if the AI invents half a work order?": for anchored tasks like this — structuring a dictation against the active equipment, order and shift — the AI's reliability is very far from free generation's hallucination problem ^[1], and even so every order stays signed by whoever dictated it and reviewable before triggering any intervention.

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