The audit dossier that builds itself

The problem

Weeks of rebuilding traceability every time an auditor or an OEM asks.

IATF audits and OEM customer claims require evidence that's rebuilt manually today, risking a non-conformance over documentation gaps.

• On IATF 16949 audit day, or when an OEM requests evidence for a warranty claim, someone spends weeks rebuilding traceability from folders, emails and photos on phones. The work falls on the same quality engineers who should be improving the process.
• Critical changes — an alloy change, a new book mold, a COS welder adjustment, a formation program update — happened, but their evidence is scattered. Nobody planned for them to be gathered into a single place later. Formation program updates and separator supplier changes are covered the same way, with their own evidence set.
• Each OEM adds its own customer-specific requirements, so the same change has to be documented in several ways. Formats differ, and keeping all of them current by hand is a full-time job nobody has.
• A gap in that documentation is enough for a non-conformance even when the change itself was done right. Auditors judge what can be shown, not what was done.

How it fits the IRIS system

Edge captures, Agents organizes — the dossier is assembled as the change happens.

Edge captures pre/post photos and data for every critical change; Agents organizes it all by standard or OEM customer-specific requirement and generates a downloadable dossier per period.

The evidence is produced at the moment of the change, not reconstructed weeks later. Edge records the before and after, and Agents files it by clause and by customer, so the dossier for the next audit already exists. Audit preparation stops being a project and becomes a download.

See the full IRIS architecture →

Before and after

Audit preparation today versus the self-building evidence pack

Aspect	Today	With iLEAN Edge + Agents
Pre-audit evidence rebuilding	Weeks of a quality engineer	Zero
Before and after proof of a critical change	Photos on someone's phone	Captured by Edge at the change
Organized by IATF clause	By hand, under deadline	Automatically, per period
OEM customer-specific requirements	Documented again for each OEM	Filed per customer from the same evidence
Answering a warranty evidence request	Days of searching	A downloadable dossier
Documentation non-conformance risk	Real at every audit	Eliminated

Weeks of pre-audit rebuilding → zero. Documentation non-conformance risk → eliminated.

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 how many audits and OEM claims you handle per year. The more OEM customers with specific requirements, the shorter the payback tends to be.
• From weeks of pre-audit rebuilding to zero. The quality engineers who used to rebuild evidence go back to improving the process.
• Documentation non-conformance risk eliminated, which protects the IATF certification. A clean audit trail also shortens the auditor's time on site, which every plant manager notices.
• And the OEM relationship, because evidence for a warranty claim is delivered in hours, not weeks. Which matters most when the OEM is deciding who gets the next battery program.

Estimated payback 6-12 months depending on audit and claim frequency, protecting the certification and the OEM relationship. *Figure to validate*.

And the fair question from the production manager

“Will an auditor accept a dossier assembled by AI?” — the dossier does not invent anything: every item is a photo, a log entry or a signature captured at the moment of the change. Classifying each item under the right clause is an anchored task, where the best models drop below 1.5% error [1], and the quality engineer reviews and releases each dossier before it goes out. Every item keeps its timestamp and its author, so the auditor can trace it back to the line. Nothing is added that was not captured at the moment of the change.

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