The FAT dossier builds itself, not at the last minute

The problem

The last gate before shipping is rebuilt by hand, unit after unit.

The FAT is the last gate before shipping mission-critical equipment; today someone spends hours per unit gathering photos, readings and signatures from scattered folders, right as unit volume climbs.

• Before a switchgear lineup or a UPS ships to a data center, the end customer requires a documented Factory Acceptance Test, often witnessed in person by their engineers and always reviewed before payment milestones.
• The evidence exists: pre and post photos, dielectric and insulation readings, functional checks, signatures. It is scattered across cameras, test sets, folders and inboxes, each named by whoever saved it.
• So someone spends hours per unit gathering it, renaming files and building the dossier, right as unit volume climbs and the FAT bay becomes a queue.
• And a missing photo or an unsigned reading is a documentation nonconformity that the customer's witness finds before you do, on the day the unit should be leaving.

How it fits the IRIS system

Edge captures each test as it runs; Agents organizes the dossier by standard and by unit.

Edge captures photo and readings per unit; Agents organizes the dossier by certifying standard (UL, ISO 9001, customer spec) and by unit or period. Zero human work after the fact.

The dossier closes while the test is being run, not the night before shipment, with every photo and reading already tied to the unit serial. Anything rebuilt afterwards inherits whatever nobody recorded, and that is exactly where a customer's witness looks. Here the same evidence serves the witness, the certifying body and the period audit, indexed three different ways.

See the full IRIS architecture →

Before and after

FAT documentation today versus the dossier that builds itself

Aspect	Today	With iLEAN Agents
Hours per unit assembling FAT evidence	Several, by hand	Zero after the test
Dielectric withstand reading	Copied from the test set	Timestamped and attached to the unit
Pre and post test photos	On a phone or a shared drive	Captured and filed per test step
Organized by UL, ISO 9001 or customer spec	Rebuilt for each request	Same evidence, indexed per standard
Missing signature or reading	Found by the customer's witness	Flagged before the unit is released
Dossier for a period audit	Days of collection	Filtered and downloaded

Hours per unit rebuilding FAT evidence → zero. Documentation non-conformity risk → eliminated.

Impact estimate

FAT dossier automation: estimated impact, to validate with your unit and audit volume.

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 unit volume and audit frequency, counting only the hours recovered in the FAT bay and in audit preparation.
• Hours per unit rebuilding FAT evidence go to zero, and the documentation nonconformity risk is eliminated because gaps are caught while the unit is still on the test bay.
• The relationship with hyperscale customers that require documented FATs is protected by a dossier that is complete every time, not only when someone had the time.
• And when the auditor arrives, the period file is a filter, not a week of collection by the quality team. The same applies when a customer asks, years later, for the FAT record of one serial.

Estimated payback 6-12 months depending on unit volume and audit frequency, protecting the relationship with hyperscale customers that require documented FATs. *Estimate to validate.*

And the fair question from the production manager

“Isn't this just another document management system?” — the difference is the direction. A document system stores what someone decides to upload afterwards; here the evidence is captured during the test and filed by unit. Reading a test-set display and indexing a reading is an anchored task, where the best models drop below 1.5% error [1], and the test engineer signs the dossier before release.

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