Roll-to-VIN traceability is earned day by day — not at 10 pm when the call-back lands.

The problem

The traceability exists — but across six different systems.

In a mid-sized airbag or seat-belt tier-2, the information linking the roll to the VIN exists — but it lives scattered: the yarn supplier's certificate in a PDF in the purchasing manager's folder, loom reports in the MES, roll scans in a spreadsheet kept by the cutting supervisor, module lot in the ERP, the tier-1's purchase order number in emails, VINs in the system the tier-1 sends back days later. When an OEM call-back arrives, someone has to reconstruct the graph by hand. And reconstructing is slow, expensive and prone to the kind of errors an IATF 16949 auditor spots.

The three realities almost nobody manages to cross-reference on the floor in time:

1. Which yarn lot is in this particular roll — including mid-roll lot mixes, which happen often.
2. Which cuts and which modules came out of that roll — and where the defective meters were discarded.
3. Which VINs each module ended up in — data that comes back from the tier-1 days or weeks later.

The classic system (chaining it by hand when the tier-1 asks) works 99% of the time. The remaining 1% is the call-back widened as a precaution to 10x the VINs actually affected — because the boundary could not be proven with data in time.

How it fits the IRIS system

iLEAN Tracer does not replace the quality manager — it keeps the dossier permanently assembled for them.

The roll-to-VIN traceability problem is not a lack of knowledge: the quality manager knows exactly what to cross-reference. The problem is speed and volume: cross-referencing thousands of rolls and modules a month against thousands of delivery notes, reports and emails, while keeping the graph always ready. iLEAN acts as the putty between the six systems where traceability lives on islands.

Connect captures everything that comes in (delivery notes, tier-1 emails, scans). Tracer chains roll-cutting-module and keeps the graph. The person signs every file that goes out to the tier-1 — the AI does not.

The iLEAN pieces applied to roll-to-VIN traceability:

• iLEAN Tracer (Agent) — a specialized agent from the iLEAN Agents family; it lives in Central and works 24/7. It maintains the yarn lot → loom → roll → cutting → module → tier-1 → VIN graph. When the tier-1 asks for data, the graph already exists. When a call-back arrives, the agent returns the affected VINs bounded — not estimated upward as a precaution.
• Connect — captures everything: yarn delivery notes, supplier certificates (PDF), loom reports (MES), QR/RFID scans in and out of cutting, module lots (ERP), tier-1 emails with purchase order numbers and returned VINs. Email, WhatsApp, the cutting panel, paper: it all comes in and becomes graph.
• iLEAN Edge — wherever it is needed, at the QR/RFID scan of the roll entering cutting or of the module leaving the line. Local reading, asking nothing of the old MES.

See the full IRIS architecture →

Before and after

Traceability reconstructed by hand vs. a live file with iLEAN Tracer

Aspect	Manual reconstruction when needed	With iLEAN Tracer
State of the graph	Exists, chopped up across 6 systems	Live and unified in Central
Response to an OEM call-back	Hours or days reconstructing	Minutes: the graph already covers it
Affected VINs	Widened as a precaution	Bounded with data
Mixed lot in a single roll	Assumption: all one lot	Transition meter recorded
Operation with no network on the floor	n/a	Edge keeps scanning QR codes
File for the IATF 16949 auditor	Prepared in the preceding weeks	Generated on the spot, with signature

Impact estimate

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

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

• Mid-sized airbag/seat-belt tier-2: 4-12 looms, 1-3 cutting lines, 1-2 sewing/module lines, supplying a European or American tier-1 under IATF 16949.
• Tracer pilot (agent + Connect capturing suppliers and tier-1 + Edge QR/RFID scanning across roll-cutting-module). First value expected within a few weeks: the first roll-to-module file chained on its own.
• Indicative payback between 4 and 9 months, depending on monthly module volume, audit frequency and historical exposure to nonconformity or call-backs.
• Reduction in response time to a tier-1 request in the order of ≥30% already in the pilot — plus a lower risk of a recall being widened for lack of data in time.
• The hard lever is twofold: recurring savings on the pre-audit office work, and a lower cost for any recall that does arrive, by bounding the VINs actually affected.

And the quality manager's reasonable doubt

“What if the agent makes up a roll-to-module association?” — the agent only recontextualizes data that already exists. It extracts lots from real PDFs, chains real QR scans, associates reports and modules by their identifiers. It does not generate identifiers or VINs. Hallucination is a problem of free generation, not of anchored tasks: the best models brought the error below 1.5% on tasks anchored to the source ^[1]. And even so, the person signs every file that goes out to the tier-1 — the AI does not.

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