One piece flow in electronics with AI — the single-piece flow your SMT line never managed to sustain.

The problem

Single-piece flow does not forgive, and in high-mix electronics it breaks every single day.

One piece flow is the Lean principle by which a part moves from one station to the next one at a time, with no buffers between stations. It works in the book and rarely on a real SMT floor. The reasons:

1. Micro-stops are invisible to the MES. The operator takes 90 seconds to swap a feeder, 40 to check the BOM of the new ECN, 2 minutes to ask quality to look at a solder joint. The MES still reads "in production" — flow is already broken.
2. High mix multiplies changeovers. Ten part numbers a day means ten vulnerable windows at every station: component change, reflow program change, screen change. Every change is an opportunity to drift out of sync.
3. WIP between stations is self-defence. When a station is left waiting because the one before it stopped, the operator builds work ahead and creates a private buffer. Nobody forces it: it is done to avoid standing idle. That is where the bin is born.
4. And defects surface late. By the time AOI flags a defect in the batch, that component position already has 200 PCBs behind it. Rework is expensive and root-cause follow-up means digging through five screens.

It is exactly the two-gaps syllogism applied to SMT flow: the data exists (every station, every feeder, every AOI knows it), but it lives on islands; and even if you joined it all up, there was no agent operating it in real time to anticipate the drift. AI has just delivered both missing pieces.

How it fits the IRIS system

iLEAN does not replace AOI or ICT — it glues them together and connects them to the live BOM.

The modern SMT plant is already full of good vertical software: AOI, ICT, an electronics MES, an ERP. Each does its own job well and almost none of them talk to the one next door in real time. iLEAN acts as the filler that seals those cracks, without asking you to replace any system.

Edge watches the station. Connect cross-references the live BOM and the day's ECN. The agent flags the drift before the bin forms. The person decides the adjustment — the line does not reorganize itself.

The three iLEAN pieces applied to one piece flow in SMT/PCB:

• Edge — a terminal with vision and/or signal reading at every critical station. It measures effective cycle, micro-stops and cosmetic defects ahead of AOI. It reads the operator screen when direct integration is not available. It works without a network: if the plant loses connectivity, Edge keeps logging cycles and resuming local synchronization.
• Connect — captures the live BOM from the MES, the ECNs arriving by email from the customer, the process owner's WhatsApp message with the reflow profile adjustment. A BOM change no longer reaches the station half an hour late through forwarded emails: it arrives at second zero.
• Agent — cross-references cycle per station with AOI, ICT, BOM and ECN. It detects the first station drifting and proposes an adjustment (relief, feeder change, program review). If AOI flags a recurring defect, it prepares a root-cause dossier with photo, chart and proposal. The engineer signs off — the agent does not stop the line.

See the full IRIS architecture →

Before and after

Traditional one piece flow vs. one piece flow with iLEAN

Aspect	Classical SMT line + AOI/ICT	With iLEAN Edge + Connect + Agent
Micro-stop detection	Only if the operator logs it by hand	Automatic per station, in milliseconds
Station-to-station synchronization	By the line leader's eye	Tracked continuously by the agent
Reaction to a new ECN	Email plus paperwork plus a per-shift update	Connect delivers it to every station in seconds
Recurring AOI defect	Digging through 5 screens for root cause	Per-batch dossier with photo, chart and proposal
WIP between stations	Operator self-defence — a permanent bin	Trending towards real single-piece flow
Traceability file per PCB	Rebuilt by hand if a claim arrives	Automatic, with feeder, ECN and AOI history

Impact estimate

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

The block below is an estimate to be validated against the actual data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• SMT line with 6-10 critical stations, high mix (8-15 part numbers a day), AOI + ICT + electronics MES already installed.
• Edge pilot on the 2 bottleneck stations + Connect against BOM/ECN + a synchronization agent. First value expected within a few weeks: the first micro-stop map per station, already discussable with the process team.
• Work-in-process between stations cut ≥ 30% after the first quarter. Internal lead time cut ≥ 30% on the pilot line. Hard levers: working capital released plus delivery time to the customer.
• Indicative payback between 4 and 9 months, depending on the hourly cost of the bottleneck and on how critical lead time is for your end customer.

Why this works now

The quality bar in automotive-grade electronics is counted in parts per million, not in percentages, and it only holds up with in-line detection and a fast closing loop. An AI model asked to opine in the abstract will fill the gaps; the same model anchored to verifiable data — measured cycles per station, the live BOM, AOI and ICT results, real feeder history — stops guessing and starts answering ^[1]. iLEAN does not invent the standard: it makes it reachable with the AOI/ICT you already own, and leaves the new capability inside your process team.

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

Keep pulling the thread: Cellular manufacturing with AI · Jidoka high mix low volume · All Lean methods