Inovai Platformrunning Testing, monitoring and evidence in one placeIt attacks your AI after every change and documents every finding. How the platform tests

Inovai / Case study

At Favex, certificates were retyped by hand.

Our AI extracts the certificate, checks it against the standard, and the operator handles only the exceptions.

Problem

Hundreds of certificates a month, around 130 suppliers, 12 languages, hours of routine work.

What we’ll do

Extraction, limit checks against the standard, exceptions to the operator.

Result

10,000+ certificates in production, 28 s per extraction, every value with its source.

How it works

Four steps; a person decides on the exception.

deviation
Step 1IntakeEmail, upload or a whole folder. A PDF with several certificates splits itself.
PDFAIHeats
Step 2ExtractionChemistry, mechanical properties, tests and notes from tables, in 12 languages.
C 0.17≤ 0.20Mn 1.42≤ 1.60P 0.031> 0.025
Step 3Check against the standardLimits by steel grade and thickness, including exceptions from the notes in the standard.
1exception11confirmed in one go
Step 4Confirmation and exportThe operator confirms certificates without warnings in one go, and the data goes straight to the ERP.Verified by a human

What changed

The operator checks, not retypes.

No more retyping

Certificates without warnings pass with a single confirmation. Attention goes only to the exceptions.

Every value has a source

Each decision shows what was in the PDF, what was in the standard and who confirmed it.

The system learns from corrections

Repeated corrections of the same field make the next extraction more accurate.