The certification dossier, ready without manual rebuild

The problem

The evidence exists the day the cooler is tested. It is rebuilt the week before the audit.

The day the product certification auditor shows up, someone today spends weeks reconstructing which units from which lot passed which test with what result. The auditor asks for evidence, the plant shows scattered records, and a documentation gap can cost the certification of an entire model line.

• The day the product certification auditor arrives, someone has spent weeks reconstructing which units from which lot passed which test, with what result. It is usually the quality engineer, pulled away from the line for days to search folders and spreadsheets.
• The evidence is scattered: test sheets, lab spreadsheets, photos on phones, records in the ERP. Each source was built for a different purpose, and none of them was built for an auditor.
• The auditor asks for evidence and the plant shows scattered records. A documentation gap can cost the certification of an entire model line.
• And with UL, NSF and ENERGY STAR each asking for their own view of the same tests, the rebuilding happens once per certifier. The evidence is the same; the indexing is different, and so the work is repeated.

How it fits the IRIS system

Edge captures each final test; Agents files it by certifier, model and period.

Edge plus Agents: every final test automatically generates an auditable dossier — a photo of the measurement result, a timestamped decision log, the technician's signature — that Agents organizes by certification standard and by period/model, with no human work after the fact.

The dossier is not built for the audit; it is closed at the moment each cooler is tested. What Agents does is present the same evidence in the shape each certifier asks for, without anyone rebuilding it. Every gap is found the day it happens, when it can still be completed, instead of the day the auditor asks for it.

See the full IRIS architecture →

Before and after

Pre-audit rebuilding versus the dossier that keeps itself

Aspect	Today	With iLEAN Agents
Preparing for a UL or NSF audit	Weeks of rebuilding records	Zero rebuilding, a review
Evidence per final test	Scattered across sheets and files	Photo, decision log and signature
UL, NSF, ENERGY STAR views	Rebuilt for each certifier	Same evidence, filtered by standard
Missing test record	Found by the auditor on site	Detected when the test happens
Model line certification	At risk from a missing record	Backed by every unit tested
Who assembles the dossier	Quality, by hand	Agents

Weeks of pre-audit record rebuilding → zero. Documentation non-conformity 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 audit frequency. The more certifications and model lines the plant carries, the larger the saving.
• Weeks of pre-audit record rebuilding go to zero.
• The risk of a documentation nonconformity is removed, because a missing record is detected the day the test happens. A model line's certification no longer depends on whether one sheet was filed correctly.
• And the quality team stops preparing the same evidence three times for three certifiers. Quality engineers return to improving the product instead of reconstructing its history.

Estimated payback of 6-12 months depending on audit frequency. Estimate to validate.

And the fair question from the production manager

“Isn't this just a document management system with a new name?” — no: a document system stores what someone uploads afterwards, with the same gaps. Here the evidence is captured at the test itself. Filing each record under the right certifier, model and period is an anchored task, where the best models drop below 1.5% error [1], and the quality manager reviews the dossier before it goes to the auditor. Nothing is sent to a certifier without that human review, and every record keeps a link to the original test photo.

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