The process engineer dictates at the press, and the PFMEA stays alive

The problem

The PFMEA is written when the process is launched — and then dies in a spreadsheet.

In a Tier-2 metal stamping plant, the process engineering lead is not sitting in front of a screen: they are walking the plant — press hall, die room, metrology lab, shipping — with their hands full, 6-7 hours a day. They are the person who first sees the patterns no automatic control reports:

• An incipient crack showing on a drawn part — a failure mode the lead spots at the first-off long before PPM appears in the Tier-1 customer report.
• A new burr that was not there before — "since this lot the die is leaving an edge on operation 30" — which nobody notes down at the time because there is nowhere to type on the shop floor.
• The effect of a steel change from a different heat — real technical knowledge, with its mill certificate behind it, reconstructed days later from memory, if it gets reconstructed at all.

The part number's PFMEA was written when the process was launched and then frozen in a spreadsheet. Real line events never go back into it: they live in the lead's head and die at the end of the shift. They do not enter the system, they are not cross-referenced with live data and, above all, they do not update the failure modes or feed the plant QC-Story. Process engineering "does not find out" until PPM appears in the Tier-1 report — and by then people are fixing symptoms, not causes.

How it fits the IRIS system

Connect in hands-free voice mode — the lead takes notes by dictating, without stopping.

The earpiece is not a system that records the lead throughout the shift: it is a tool they activate deliberately when they want to note a technical incident on their walk through the gemba. They decide when to speak and what to dictate. Connect structures it, ties it to the active plant context and returns it within seconds.

The lead says "iLEAN, take a note". They dictate the technical incident in their own words, without stopping. Connect structures it as an incident tied to the active lot, die and press. Seconds later the response arrives by voice through the earpiece, or as a push to the line manager's tablet.

How Connect works in voice mode in a Tier-2 metal stamping plant:

• A personal wake word — the lead opens the dictation with their own phrase, one only they say deliberately. Outside that phrase, the Bluetooth earpiece transcribes nothing and sends nothing.
• Free dictation, no forms — they speak in their own words, exactly as they would tell a colleague: "incipient crack on the operation 30 draw, lot from the new heat, have quality run a first-off before continuing".
• The LLM structures the incident — Connect turns the dictation into an incident with clear fields, automatically tied to the lot, the die and the press active at that moment, without the lead having to recite any of it.
• Response by voice or push — the confirmation arrives through the same earpiece or, if more context is needed, as a push alert to the line manager's tablet. The lead keeps walking, with the incident already in central memory.
• Enabler of the IATF cycle — every dictated incident feeds back into the living PFMEA of the part number and the plant QC-Story: continuous improvement stops running on anecdote and starts working on facts from the gemba.

See the full IRIS architecture →

Before and after

Knowledge that dies at the end of the shift vs. information that enters the system in real time

Aspect	Without an earpiece	With iLEAN Connect in voice mode
Technical incident spotted on the floor	Stays in the head until the end of the shift, if it makes it	Dictated in the moment, hands free, without stopping
Link to lot, die and press	Depends on somebody remembering and reconstructing it	Automatic — the system already knows what is active
The part number PFMEA	Written at process launch and frozen in a spreadsheet	A living PFMEA that receives real line events and updates failure modes
A new crack or burr on a die	Spotted once PPM shows up in the Tier-1 report	Noted and tied to the die the first time the lead notices it
QC-Story and problem analysis	Start by reconstructing the case from memory and fix symptoms	Start with real gemba context and attack causes
Cross-reference with live line data	Does not exist: human observation and data live apart	The incident is cross-referenced in real time with live plant data

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.

• Tier-2 metal stamper with a process engineering lead who walks the gemba 6-7 hours a day, hands full and with no practical access to a keyboard between presses, dies, metrology and shipping.
• Connect voice pilot on one line or one part number — without changing the lead's role and without adding administrative tasks.
• The main return is not an accountable saving in hours: it is that the PFMEA, the QC-Story and every problem analysis start with real gemba context instead of memory-based reconstructions in a meeting.
• There is an expected operational return too: early detection of die failure modes — the crack or the burr the lead notices days before PPM shows up in the Tier-1 report. That improvement is an estimate to be validated with the plant's continuous improvement lead, not a generic figure invented here.

And the fair question from the plant director

"Does this not end up as a system that records everything my process lead says?" — no. It is exactly the reverse: the earpiece hears nothing until the lead activates it with their wake word. There is no background recording, no continuous transcription, nobody listening to their conversation with a press operator. The only thing that enters the system is what they themselves choose to dictate — because they want it on record, not because somebody above asked for it. And the second question, "what if the AI makes half an incident up?": for anchored tasks like this one — structuring a dictation against the active lot, die and press — the AI operates far from the free-generation hallucination problem ^[1], and even so every incident is signed by whoever dictated it and is reviewable before it feeds any PFMEA.

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