Audit evidence, generated at the moment it is produced

The problem

The non-conformity is raised on what you cannot prove, not on what went wrong.

In automotive, evidence is not an extra: it is part of what you sell to the auditor. Sector standards and customer audits require you to demonstrate control over special characteristics, over changes and over deviations. The problem is that every standard wants the same information in a different shape. Today that restructuring is human work, done after the fact and under pressure — hunting for first-off records in folders, printing lab reports and explaining why the record from a night shift four months ago is missing. And that is exactly the scenario where gaps appear. The non-conformity is almost never raised because the process was wrong: it is raised because you cannot prove it was right.

• IATF 16949 and the customer-specific requirements ask you to demonstrate control over special characteristics, over changes and over deviations. Evidence is part of what you sell to the auditor.
• Every standard and every OEM wants the same information in a different shape, and that restructuring is human work done after the fact and under pressure.
• Hunting for first-off records in folders, printing lab reports, explaining why the record from a night shift four months ago is missing: that is audit preparation today.
• And that is exactly where the gaps appear. The non-conformity is almost never raised because the process was wrong; it is raised because you cannot prove it was right.

How it fits the IRIS system

Edge and Connect capture at the moment it happens; Agents organizes by standard.

Edge captures visual evidence at the moment it is produced — before and after photos of the critical points at every changeover, the press process curve, the weld curve and the engraved serial number, lab results, rejects with their images — each with signature and timestamp, all tied to the part's serial number. Agents maps every piece of evidence to the requirements of each standard and builds the pack by period, by part number or by serial number. And it proactively flags gaps — "that mold's night shift is missing the validation of one point" — while they can still be closed, not in audit week. The pack is signed by quality, not by the AI.

The gap is flagged the day it happens — that mold's night shift is missing the validation of one changeover point — while it can still be closed, not in audit week. And the pack is signed by quality, not by the AI. The same evidence, re-indexed, serves the certification body and every OEM with its own requirements.

See the full IRIS architecture →

Before and after

Today's audit preparation versus the permanent pack

Aspect	Today	With iLEAN Agents
Preparation before an audit	Person-days of reconstruction	Zero
Changeover evidence	A signed checklist, if found	Before and after photos per point, timestamped
Weld curve on the airbag tear seam	Inside the cell	In the pack, per serial number
Documentation gap	Found in front of the auditor	Flagged the same day
Answering an OEM complaint	A week	Hours, with serial-number evidence
Same evidence, another OEM	Rebuilt in full	Re-indexed to that customer's requirements

from person-days of document reconstruction before every audit to zero · from finding the gap in front of the auditor to seeing it the same day it happens · from a customer complaint that takes a week to answer to an answer with serial-number evidence in hours.

Impact estimate

Impact estimate — to be validated with your 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.

• Estimated payback 6-12 months, depending on audit frequency and the number of OEMs with their own requirements.
• The person-days spent preparing each audit — hunting first-off records, printing lab reports, reconstructing a night shift — saved directly.
• Protection of the IATF 16949 certification and of standing in customer supplier panels — a loss measured not in euros but in programs you do not win.
• And a customer complaint answered in hours with serial-number evidence instead of in a week.

protection of the certification and of standing in customer supplier panels — the loss of which is not measured in euros but in programs you do not win — plus the direct saving of the person-days spent preparing audits. Estimated payback 6-12 months depending on audit frequency and the number of customers with their own standard. *Estimate to be validated.*

And the fair question from the production manager

"Isn't this just another document manager?" — the difference is direction. A document manager stores what somebody decides to upload, after the fact, and inherits the same gaps. Here the evidence is captured at the moment it is produced, each item with its signature and timestamp, and Agents maps it to the requirement of each standard. Where it reads a record rather than a sensor, that is an anchored task where the best models drop below 1.5% error [1], and quality signs the pack.

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